Think of the Children

Cancer diagnostics and treatment have come a long way over the past 50 years. Before the advent of chemotherapy, surgery – sometimes aided by radiation therapy – was the only effective intervention for cancer patients. Although this approach has long been the primary treatment modality for adults, where the goal is to remove the entire tumor, surgery is not as effective in younger patients. Childhood cancers, even those with apparent similarities, are distinct from those seen in adults – with different developmental stages, tissues of origin, mutations, and gene fusion events.

With the discovery by Sidney Farber that a folate antagonist improved survival in children with acute lymphoblastic leukemia came new hope in the fight against pediatric cancers. The application of chemotherapeutic agents quickly spread across all childhood tumors – and it became clear that children typically respond better to chemotherapy than their adult counterparts. Survival rates for pediatric cancer patients – which until this point had been far inferior to adults – began to rise and even surpassed those for adult cancers.

More recently, the field of immuno-oncology has begun to flourish – but, despite showing promising signs in the adult cancer world, immunotherapy is less effective in pediatric cancers. The efficacy of immunotherapy depends on the strength of the patient’s own immune response – often absent in childhood cancer patients with few tumor mutations. More encouraging is the use of therapies targeting unique gene fusion events. These events are common in pediatric cancers, clearly differentiating their molecular pathogenesis from that of adult disease.

As precision medicine continues to serve patients of all ages, it is clear that our understanding of the genetic drivers behind childhood cancers will be key to future progress and more finely tuned therapeutic interventions. Two initiatives aiming to improve outcomes for childhood cancer patients are the OncoKids Cancer Panel and the Pediatric Cancer Genome Project. We spoke to two of the experts working at the forefront of pediatric laboratory medicine to learn more about the cutting-edge molecular approaches to childhood cancer diagnosis and treatment.

A World Apart

Childhood and adult cancer drivers are distinctly different – and the approaches must be equally specialized

An interview with Timothy Triche

What led you to pediatric laboratory medicine?

I originally trained in surgical pathology, but I was also interested in research and chose to do a National Cancer Institute (NCI) fellowship at the National Institutes of Health. I thought this would be a brief stint and then I would move on – but, once I arrived, I discovered that they had no full-time pediatric pathologists. I was asked to present tumor boards and review cases with the pediatric oncologists and surgeons, which started as a sideline, but soon became my main interest. There was such a wealth of interesting cases arriving at the NCI that, eventually, I focused exclusively on pediatric cases. Although I still have an interest in adult cancer, the vast majority of my time for the last few decades has been spent on pediatric oncology, and I have expanded from surgical pathology into many other areas.

Tell us about the OncoKids Cancer Panel.

While working at the NCI, it quickly became obvious that the adult and pediatric cancer cases I signed out side by side were very different from each other – not just in terms of the diagnoses, but also their characteristics; for example, where tumors occurred and how they behaved. Why were they so different? We didn’t know at the time that the drivers of pediatric cancer are completely different from those of adult cancers. There is very little overlap because most adult cancers are the result of accumulated mutations in the genome; however, we knew early on that this wasn’t the case for childhood cancers. We started to search for features that could explain pediatric cancers and noticed recurring chromosomal breaks across various tumor types. We realized that these breaks – where a piece of one chromosome fuses with another – were likely giving rise to driver genes, and we eventually characterized them as fusion genes. These fusion genes are extremely common in childhood cancer and, although they sometimes occur as a secondary issue in adult cancer, they are frequently a primary feature in pediatric cases.

When the NCI-MATCH (Molecular Analysis for Therapy Choice) Trial was announced in 2015 to assess the efficacy of targeted therapy for patients with specific gene mutations, I realized that it would miss most of the important features of pediatric tumors. The main types of cancer in adults are lung, colon, breast, and prostate – all carcinomas that arise from lining or covering tissue. In contrast, childhood cancers ultimately derive from the mesoderm or neurectoderm; there is essentially no carcinoma. I wanted to create a cancer assay that reflected the tumors that occur in children, adolescents, and young adults, so we at Children’s Hospital Los Angeles (CHLA) spent several years developing the OncoKids panel with Thermo Fisher Scientific. Working with pathologists, oncologists, surgeons, and industry experts, we assembled the content required for pediatric cancers. Although some of the assay’s DNA content remained the same due to an overlap between mutations in children and adults, the RNA content is unique. There are approximately 1,400 different combinations of potential fusion genes in childhood cancers, which we reflected in a panel heavily skewed toward RNA content.

How is it used?

The panel is particularly useful because it can i) identify inherited components, which account for around 15 percent of childhood tumors; ii) establish a precise diagnosis, which is crucial when introducing potentially toxic chemotherapeutic agents; and iii) detect specific mutations that can be treated with targeted therapeutics. For example, relatively early on in our use of the panel, a very young patient presented at CHLA with a life-threatening neck tumor – but the original biopsy interpretation was inconclusive. We were asked to carry out a rapid analysis using a small amount of formalin-fixed, paraffin-embedded tissue, which is historically difficult to assess. We processed the tumor material and identified an unusual childhood cancer gene fusion – tropomyosin fused to tyrosine kinase – providing us with a target for therapy. Working with the head of our solid tumor service, we deployed an NTRK inhibitor that same day, and the patient responded very favorably. The tumor became undetectable over a period of weeks and, with no further treatment, the patient has remained disease-free for two years. This case demonstrates the importance of knowing exactly what you’re treating and selecting the appropriate agent.

The RNA content in the panel is crucial because gene fusions often establish an unequivocal diagnosis and determine how the patient will be treated and whether there is a suitable targeted agent available. We now present these results in three different weekly or biweekly tumor boards: one for brain tumors, a second for leukemia and lymphoma, and a third for solid tumors. The molecular pathologists discuss the findings from the OncoKids panel in conjunction with the surgical pathologists – and that comprehensive diagnostic workup is then discussed with the treating oncologists.

What are the main differences between approaches for children and adults?

Before the advent of chemotherapy, the only effective treatment for cancer was surgery. The survival rate for childhood cancer in that era was abysmal because surgery rarely works by itself in these cases; pediatric tumors tend to move around and often become metastatic early in the course of disease. However, the outlook changed with Sidney Farber’s discovery that drugs could cure leukemia. That knowledge rapidly translated to all other childhood malignancies – and it was quickly established that children respond far better to chemotherapy than adults. Treatment regimens in children quickly became multimodal affairs whereby surgical excision was followed by chemotherapy and radiotherapy – and survival rates improved drastically.

Another key difference is the effectiveness of immunotherapy, which relies on the development of a host immune response that the tumor suppresses. Although adults often respond well to immunotherapy, childhood cancers aren’t usually associated with tumor burden and tumor mutations. As a result, immune response in childhood cancers is nowhere near as robust, and immunotherapy is reserved for rare pediatric cases with a significant amount of mutation. That leaves us still very much dependent on classic chemotherapeutic agents – but our hope is that we will see an increase in less toxic agents that target the gene fusions common in childhood cancer.

How has COVID-19 affected your work?

Many of the sensitive next-generation sequencing (NGS) technologies that we are accustomed to using for pediatric cancer suddenly became very useful, especially when dealing with poor-quality nasopharyngeal swab specimens to detect the presence of SARS-CoV-2. Specifically, we have been able to sequence viral isolates to determine whether infected individuals have transmitted the virus between themselves or acquired it separately in the outside community. It is through this testing that we spotted a viral strain called D614G, a variant of COVID-19 that was detected in Europe and spread to the east and west coast of the US. It’s not uncommon to see multiple slight sequence variants from a single isolate – and this issue of SARS-CoV-2 genomic stability is both fascinating and crucial for the development of any potential vaccine. We will continue to monitor these variants for any changes in the amino acid sequence that could, in turn, alter the spike protein and affect antigenicity.

Although the severity of COVID-19 was originally assumed to be worst in adults, the reality is that children – particularly the very young – can also have an extraordinarily severe response. In some pediatric cases of infection by SARS-CoV-2, we see a unique host autoimmune response that results in cytokine storm and widespread tissue damage, producing rashes on the skin and, in some cases, multiple organ failure. There are many hypotheses as to why we only see this response in children. My own suspicion is that many older patients have mild or asymptomatic disease because they already possess partial immunity and can rally old memory B or T cells to fight infection. In contrast, young children are being exposed to these viruses for the first time – and, although most immune systems can produce an appropriate response, some launch a destructive response that is not attuned to the viral infection. We still need to learn a lot more about this type of autoimmune disease and the potential to treat these young patients with cytokine inhibitors.

How has pediatric medicine changed over the course of your career?

I distinctly remember seeing young leukemia patients and realizing that I would need to have a very difficult conversation with the parents about the awful prospects their child faced. In those days, leukemia was largely incurable and the mortality rate was nearly 100 percent. However, over the course of a decade after the arrival of chemotherapy, we flipped to having an almost 100 percent survival rate – a phenomenal change in the outlook for childhood cancer patients.

Traditionally, clinical presentation is a cluster of symptoms and findings that we label and treat accordingly. If there is one overriding theme over the course of my career, it’s that we’ve come to appreciate the value of personalized medicine. Every patient possesses unique features against which we can develop finely tuned therapeutic interventions and management. And that’s one of the major reasons for our improved outcomes; we now recognize categories, understand genetic drivers, and instigate patient- and disease-specific treatments. We’ve also learned that, although pediatric cancer types can occur in patients aged two or 22, the behavior of those cancers will be extremely different. We now take into account age, the specific disease evolution, and any polygenic influences, all of which provide pediatric medicine with the opportunity for precise diagnosis and treatment.

Where will the field go over the next 10 years?

The personalized approach will become increasingly evident as the precision and accuracy of our tools improve. There are clearly many questions to be answered – and, although it seems as though every answer begets 10 new questions, we’re also developing the tools to analyze more and more data, variables, and features. Some interesting possibilities are emerging, especially in terms of tools that aggregate vast amounts of data beyond the capability of any one person. Machine learning and artificial intelligence now facilitate worldwide databases containing extensive information about all kinds of unique cases. We can use these databases to compare and contrast patients’ features with cases across the globe, ensuring we make informed decisions about diagnosis, treatment, and management.

As this becomes increasingly widespread, we’ll be spending more time in front of our computers trying to make sense of it all. I think these tools will become dominant features in medical management for patients of all types in the future – especially because that intelligence database can help us understand the results of our increasingly complex diagnostic assays and workups. As a result, to thrive in pediatric – and all – medicine in the future, you will have to feel comfortable working with your computer and quantitative analytic methods.

What advice would you give to anyone considering a career in pediatric medicine?

Anyone entering the field today will benefit enormously from basic training in math and statistics so that they can appreciate the use of new tools and draw conclusions from them. Because I wanted to become a particle physicist early on in my own career, I developed an appreciation for analytic methods and mathematical analysis. I never really envisioned these skills being useful in my journey through pediatric laboratory medicine – but, as the analytical tools in our armory grew, it became increasingly obvious that quantitative analysis is crucial for interpreting data that directly relates to a patient’s diagnosis. For example, we often work with biomarker profiles that provide multiple variables, which can be extremely powerful predictors if you are comfortable aggregating and analyzing the important features.

It’s also important to understand the limits, as well as the capabilities, of analytic methods. Laboratory values are frequently numeric and you have to decide when certain values are outside the normal range. In my opinion, virtually every aspect of pathology and laboratory medicine is ultimately influenced by the need to understand and use quantitative analytic methods. My advice to those considering a career in this space would be to develop these skills as part of your overall medical training.

Timothy Triche is the Co-Director for the Children’s Hospital Los Angeles (CHLA) Center for Personalized Medicine, USA.

Strength in Numbers

Data sharing is critical to gathering detailed information on rare pediatric cancers

An interview with Jinghui Zhang

How did you get into pediatric laboratory medicine?

I joined St. Jude Children’s Research Hospital in 2010 after working at the NCI. At the time, The Cancer Genome Atlas Program was the center of attention – everyone seemed to be working on this project focused on adult cancer. In 2010, St. Jude invested US$65 million in the Pediatric Cancer Genome Project, created in collaboration with Washington University School of Medicine, to sequence the paired tumor-normal genomes of 600 pediatric cancer patients. Its goal was to define the genomic landscapes of some of the least understood and most challenging childhood cancers. And it was one of the main reasons I transitioned from the adult cancer world to pediatric cancer.

Another of my motivations stemmed from the 20 years of genomic data I had been studying throughout my career. When I was at the NCI, I analyzed a subset of pediatric leukemia patients and discovered an activating mutation in a kinase gene called JAK2. We started discussing how to apply this finding to clinical trials to discover JAK inhibitors that could potentially treat patients – and that was the moment I began thinking about working in pediatric oncology. I thought, “What could be more rewarding than research that leads to improved clinical outcomes in childhood cancer patients?”

What is the value of precision medicine for pediatric cancers?

Precision medicine has existed for a relatively long time in the context of pediatric cancer; however, in the past, we used techniques such as cytogenetics and polymerase chain reaction testing to find biomarkers and classify cancer subtypes. With next-generation sequencing (NGS), we can classify patients much more precisely and effectively. For example, in pediatric leukemia, we can look at gene fusion patterns to define low- or high-risk patients and apply chemotherapy according to their global genomic profile. Another great benefit of precision medicine is the range of treatment options we can offer patients with specific mutations. We can search for genes such as JAK2 and apply targeted therapy if there are activating mutations or gene fusions in the kinase. For patients with vulnerabilities, such as high mutational burden or DNA mismatch repair, NGS can assess potential response to immunotherapy – and, in solid tumors, we can ask whether the patient is susceptible to PARP inhibitors.

But the aspect of precision medicine that is most specific to the field of pediatrics is cancer susceptibility. In 2015, we performed a comprehensive analysis of germline inherited susceptibility to pediatric cancer because of the uniqueness of our patients with cancer at a very young age. We can discover pathogenic mutations in the germline genomes of pediatric cancer patients – and include family studies to analyze inheritance patterns between relatives. We can also monitor patients for potential secondary cancers further down the line, and potentially intervene with preventative procedures to decrease the chance of recurrence during adulthood.

What is St. Jude Cloud and why is it so important?

A really challenging aspect of working with pediatric cancers is that they are such rare diseases. For certain subtypes, we sometimes find that there are fewer than 200 cases in the USA, which often means insufficient access to samples and not robust correlative analyses between genomic profile and clinical outcome. In terms of sharing the information we do have on rare pediatric cancers, the predominant model at the moment is to upload data to an online repository that others can access, download, and share. However, this approach is not efficient, because every individual user must download their data of interest, which requires resources and time. It’s also not sustainable for institutions or labs that don’t have a large digital infrastructure, because the data takes up so much local space. Even for those of us with access to a high-throughput computing facility and professional support from computer scientists and software engineers, this can still be a time-consuming and difficult task.

And so, we decided to create a cloud-based infrastructure that provides users with a platform on which to upload and share data without having to download anything locally. We launched St. Jude Cloud in 2018, and the data on the platform has already been accessed by 2,500 users weekly and more than 200 researchers from 80 institutions worldwide have been granted access to raw genomic data. Although we haven’t reached the point at which all analyses can be conducted on the cloud, we believe the future lies in working entirely with cloud-based data – and that’s what we’re now working toward. This ability will be particularly useful for laboratories and institutions with limited computational resources; they will be able to integrate their own data with those already available on St. Jude Cloud and tap into this wealth of information.

The efficacy of this approach is clear – even within the first year of St. Jude Cloud, we had more users access our data than in the previous five years of uploading data to the public repository. Our users include experienced labs using the data to find new diagnostic markers, develop specific immunotherapies for pediatric cancers, investigate whether drugs effective in adult cancer can be applied to childhood cases, and even search for new cancer susceptibility genes in germline genomes. We are in the process of preparing a manuscript to demonstrate how useful the cloud-based platform can be for discovery.

Tell us more about the Pediatric Cancer Genome Project…

The initiative was introduced in 2010 by James Downing, CEO of St. Jude Children’s Research Hospital, to generate high-quality NGS data from the genome. It was important to focus on the whole genome – rather than the exome that only targets the gene-coding regions – because a high proportion of pediatric cancers are driven by gene fusions or DNA rearrangement events. Many of these DNA aberrations occur in noncoding regions, making them impossible to discover by targeting the exome alone. In all, we performed whole-genome sequencing for more than 700 paired tumor-normal genomes – and we’ve been able to identify various novel gene rearrangements. For example, we have discovered the RELA gene fusion in patients with a certain type of ependymoma and targetable NTRK fusions in a subgroup of high-grade glioma. We put a strong emphasis on studying both the germline and tumor genomes, which has allowed us to discover tumor subtypes with a high mutational burden.

Again, we felt that sharing the data St. Jude generated from the Pediatric Cancer Genome Project was crucial, not least because it was valuable to researchers around the world who are working toward improving the care of pediatric cancer patients. We made a conscious effort to publicly distribute the information, uploading the genomic data to the National Center for Biotechnology Information’s database of Genotypes and Phenotypes, and the European Genome-phenome Archive prior to publication of all studies related to the Pediatric Cancer Genome Project.

Why was it important for such a project to focus solely on children?

When you compare disease types between children and adults, you find that the most common adult cancers – breast cancer, lung cancer, and prostate cancer – almost never occur in a pediatric setting. Some of the key insights gained from adult cancers may not be applicable for pediatric cancer. Even for cancers that share the same tissue of origin, the driver mutations can be distinct between the two. For example, in many adult brain tumors, EGFR amplification is common – but, in childhood brain tumors, it’s rare; instead, we often see amplification of PDGFR mutations. Notably, the epigenetic regulation caused by histone H3 mutations is present in over 80 percent of pediatric cancers – but completely absent in adult glioblastoma.

In a 2018 study, we compared all of the different mutations in childhood cancer subtypes in a pan-pediatric cancer study with those found in pan-adult cancer studies (1). We found that half of all driver genes in pediatric cancer are not present in adult cancer. Most are involved in transcription regulation caused by gene fusion events unique to childhood cancers. This suggests that you cannot always treat a childhood cancer patient like a small adult, because the molecular pathogenesis is so different. Instead, we must study pediatric cancer independently to search for newer, more relevant targets – and that’s something the Pediatric Cancer Genome Project strives to facilitate.

What were the project’s most exciting findings?

You’ll often hear people comment on the low mutation burden of pediatric cancer. One of the big revelations of the Pediatric Cancer Genome Project is that, even though these cancers are very rare, they are highly heterogeneous and exhibit a wide spectrum of mutation burden. For example, there are diseases like infant acute lymphoblastic leukemia, low grade glioma, and retinoblastoma that only have one driver mutation, whereas others, such as high-grade glioma, have mismatch repair defects that lead to even higher mutation burdens than equivalent adult cancers.

The second big discovery was the importance of gene rearrangements, how often they occur, and how frequently they function as the driver variants in pediatric cancer. We were surprised to find that missense mutations were not the main cause of childhood cancers; instead, up to 60–70 percent of drivers were caused by rearrangements or copy-number alterations. The final thing that struck us was the interplay between germline and somatic cells. There are cases in which pathologists will classify a cancer according to its morphology – but the tumor genome doesn’t reveal anything striking. In these cases, it’s important to take a deep look at the germline DNA to find out whether there are pathogenic mutations in the germline, some of which may have a second hit in the tumor genome.

How would you describe the impact of the Pediatric Cancer Genome Project?

The US Food and Drug Administration recently released a “Relevant Pediatric Molecular Target List” (2) – and many new targets were based on genomic abnormality discovered by the Pediatric Cancer Genome Project. These targets will allow people to pursue new drugs or test the relevance of existing therapies in the context of childhood cancers. In line with our strong data-sharing vision, we always seek to collaborate, with a view to integrating information and performing robust analysis using extended cohorts. We have participated in a new data-sharing initiative launched by the National Cancer Institute – the Childhood Cancer Data Initiative – which will collect, analyze, and share data to advance the treatment for cancer in children and young adults. Such investments allow researchers to harness collective data and resources to advance pediatric cancer research.

What are your hopes for the field?

A significant challenge we need to address is how to share data not just across the USA, but globally. Different countries have different regulations – and that’s something we’re trying to tackle by engaging our global partners and collaborators in pediatric cancer. We are working toward a system that follows global data-sharing regulations, but also allows scientists to integrate their own data on the fly, which is especially useful for the rarer cancers.

Such structure is key to the future of the field – and we are investigating the feasibility of a federated data system that would connect constituent databases around the world via a computer network, allowing someone interested in a specific cancer subtype to access and assemble relevant cases from data in many locations. As researchers, this is not a problem we can solve alone. I hope that we can work with technology companies and cloud providers to enable integrated data sharing across clouds in different regions.

My biggest hope is that researchers who are interested in finding cures for pediatric cancers can collaborate and share data across the world. This kind of engagement is so crucial – especially in this field – because no country can fight the battle against pediatric cancer alone. I believe that the COVID-19 pandemic underlines the importance of global collaboration – it’s also our greatest weapon in the fight against cancer in children.

Jinghui Zhang is Chair of the Department of Computational Biology and Endowed Chair in Bioinformatics at St. Jude Children’s Research Hospital, Memphis, Tennessee, USA