Precision medicine in pediatric oncology

Personalized cancer therapy, is not a new concept, particularly in pediatric oncology. Established pediatric protocols stratify patients into different risk groups according to therapy response, i.e., prednisone response and residual disease in acute lymphoblastic leukemia (ALL). Even in the era before response stratification, patients were stratified according to age and gender, leukemic load, and biomarkers, e.g., a T-cell ALL occurring on teenage boys was stratified into a higher risk group compared to common B-progenitor derived ALL. T-lineage persists to date as an initial biomarker of risk of relapse.

Furthermore, progress in cytogenetic techniques, and increasingly in genomic analysis, determined that certain risk groups in ALL, defined for instance by chromosomal translocation or by poor in vivo response to therapy (prednisone response on day 8, monitoring of minimal residual disease with molecular markers), are associated with a worse response to therapy.

As a result, more intensive therapies, e.g., allogeneic stem cell transplantation was used in the treatment of t(9;22) positive ALL. Similar risk stratifications are used in nearly every therapeutic trial in pediatric oncology, taking into account tumor localization, extent of spread, histologic subtype and, in recent years, increasingly molecular characterization. Detection of circulating tumor cells and circulating tumor DNA (“liquid biopsies”) plays an emerging role as non-invasive diagnostic and disease monitoring tool. Liquid biopsies are now being evaluated for neuroblastoma and Ewing sarcoma [8, 9].

First in 1960, cytogenetics became crucial for the molecular characterization of cancer [10]. Very early on, karyotyping showed a difference in the number and structure of chromosomes in cancer cells compared to normal cells [11‐14]. This resulted in the characterization of translocations that have been found in leukemia as well as in solid tumors such as Ewing and other sarcomas. Moreover quantitative genomic alterations can be characteristic of certain cancers like the amplification of the super-enhancer MYCN (myelocytomatosis oncogene of neuroblastoma), a member of the MYC superfamily and of the 450 Ma old MYC interactom [15], driving most notably high risk neuroblastomas. Discovery and exploration of oncogenes have decisively influenced therapeutic concepts and laid the foundation for today’s quantum leap in knowledge and possible therapies.

Over the past two decades, basic research expanded our understanding of fundamental aspects of cancer. Malignant deregulation of cells is due to genetic changes in somatic cells through mutation, translocation, or overexpression of genes, resulting in cellular dedifferentiation, proliferation, avoidance of cell death, and survival under cellular stress. This knowledge rendered altered genes or their genetic products, proteins, and targets for therapeutic interventions. The prime example, delivering a breakthrough in therapy, was the analysis of the Philadelphia chromosome. The translocation t(9;22) results in the BCR-ABL (breakpoint cluster region—Abelson murine leukemia viral oncogene homolog 1) fusion gene. It activates the tyrosine kinase ABL, causing autonomic proliferation in affected cells. This genetic alteration exists not only in chronic myeloid leukemia, but also in childhood acute lymphoblastic leukemia, albeit in a low percentage of patients [16].

Through the application of a tyrosine kinase inhibitor, imatinib, the induction of proliferation through ABL-activation can virtually be stopped completely and leukemic cells die. Treatments with imatinib, for example, in chronic myeloic leukemia, as monotherapy can lead to long lasting molecular remission and operational cure [17]. In ABL-activated acute leukemias, however, imatinib is only beneficial when incorporated into combined cytotoxic regimes.

In pediatric oncology, as in adult medicine, many current molecular profiling programs for patients with relapsed or refractory tumors aim to sequence tumor genomes, in order to identify genetic alterations and switch off related genes with targeted therapies. In Germany, the Society for Pediatric Oncology and Hematology (GPOH) is promoting the INFORM trial. Sequencing is performed centralized at German Cancer Research Center (DKFZ) while most of the German pediatric oncologic centers take part [18]. In the USA, the National Institute of Health and the Children’s Oncology group are performing the Pediatric MATCH trial (Molecular Analysis for Therapy Choice) [19‐21]. Presently, the results of at least eight such trials have been published or presented at symposia [19, 21‐27]. Depending on the trial, between 50 and roughly 300 samples from patients with primary disease, relapse, or refractory disease were sequenced per study. Overall, a genetic alteration that can be targeted therapeutically, i.e., for which drugs already exists, could be identified in between 10 and 30% of patients [18, 19]. These results are sobering and might even be over-optimistic, since it very much depends on the definition of targetable alterations, differentiation between driver and passenger mutations, detection of the frequency of gene fusions or minor clones, drug interactions, pharmacology, and last not but not least bioinformatics [28]. Even in the presence of a targetable alteration, available drugs may substantially differ in their efficacy depending on the cellular context. We have learned, for instance, that neuroblastoma bearing activating anaplastic lymphoma kinase (ALK) mutations respond much less efficiently upon crizotinib treatment than ALK translocations in lymphoma or lung cancer [29]. Beyond that, some trials in adult medicine that have been published did not reveal an advantage over “physician’s choice,” even, when targetable genetic alterations were identified and appropriate drugs administered [4]. Finally, as long as targeted therapies do not target oncogene addiction pathways, they may well prime for resistance, raising selective pressure to bypass the targeted pathway with alternate rescue signaling [30].

Moreover the new paradigm of dichotomy between proliferation and metastasis [31] deserves consideration in this context. Downregulation of pathognomonic fusion proteins such as EWS/ETS may be associated with reduced tumor proliferation but increase of migration and consecutive metastasis: in Ewing sarcoma (ES) the level of fusion gene expression correlates inversely with a tendency toward metastatic spread [32]. We [33, 34] as well as researchers in the USA [35] have further shown that gene products, which are overexpressed in ES, may inhibit growth of the primary tumor but favor metastatic spread [36].

It is widely accepted that, in contrast to adult cancers, most childhood cancers develop as an accident of growth or differentiation and not as a result of environmental mutagenic impact. It has also been assumed that, apart from rare hereditary cancer syndromes, there is no genetic predisposition. The trials cited above have now opened a new view on malignant diseases in children, adolescents, and young adults [37]. It has been shown in six independent studies that 5–10% of patients have germ-line mutations that predispose to cancer [18, 19, 21, 24, 25, 37, 38]. This is all the more surprising because these mutations affect primarily patients from families where there is no increased susceptibility for cancer, i.e., high incidence of malignant diseases or cancer in adolescence. Thus, we have to assume that they are de novo mutations and yet to understand many consequences of these findings, particularly for targeted therapies.

Another take home message from these trials is the significant difference in the number of mutations between cancer in childhood and adolescence vs. cancer in adults. Whereas virtually all cancers in elderly patients have multiple genomic alterations, where tumor cells are often polyploid with multiple aberrations of chromosomes; most pediatric tumors however, exhibit only few mutations and genetic alterations [4, 18]. This limits the availability and use of drugs for targeted therapies. One has to take into account, however, that most of the trials carried out so far focused on identifying mutations. In a lot of cancers the aberrant expression, overexpression and deregulated activation of certain genes is causative. Genomic rearrangements in non-coding regions may lead to massive activation of oncogenes, such as GFI1 (growth factor independent 1 transcriptional factor) in medulloblastoma or TERT (telomerase reverse transcriptase) in neuroblastoma [39, 40]. We are still missing systematic functional trials addressing this issue. Only a single institution transcriptomic trial from one of the authors’ institution revealed druggable targets in all patients and a survival advantage of patients with targeted therapies [22]. Nevertheless, together these findings highlight that expression of targets on the protein level need to be verified as well as delivery of targeting drug. As an exception to the general rule of low mutational load in childhood malignancy vs. high mutational load in cancer in the old, the hypermutational load of the malignancies caused by germ line encoded mismatch repair deficiency syndromes, has provided a successful rationale for T-cell checkpoint inhibition and for the avoidance of genotoxic therapy in these young patients [41]. Thus, more experience is to be gained, and more complexities have to be understood, before precision medicine can realize its innovative potential and strengths.

The terms “precision medicine,” “personalized medicine,” and “individualized medicine” are now part of medical concepts that seek to identify and target molecular structures in many diseases, including cancer. Former US President Barack Obama launched the Precision Medicine Initiative in 2015 (“Cancer Moon Shot Initiative”), comparing it to the first moon landing. The US National Institutes of Health are using this initiative to form new strategies for diagnosis and therapy, particularly of cancer [7].

A hallmark of cancer is the activation of genes forcing cells into proliferation or maintaining survival under stress while blocking differentiation and cell death [6]. Normally, cells receive external signals that are transmitted into the cell by receptors, for instance tyrosine kinase receptors transmitting signals through phosphorylating tyrosine residues of proteins. Deregulated activation of tyrosine kinases is characteristic of most cancers [42] (Fig. 2).

However, identification of the specific kinase in each case may not be easy, as deregulated activation can arise not only from activating mutations but also from inactivating mutations in suppressors (Table 1). The identification of tyrosine kinase inhibitors, initially in adult cancers, has provided a spectrum of substances that can be used therapeutically, as shown with imatinib in Philadelphia chromosome positive ALL and CML. In most cases, however, these drugs are effective only for a short time (with the exception of CML) when used in in monotherapy (cf. 2.2.) [30].

There are interesting examples of genetic alterations that were initially identified in different adult cancers and are now being targeted in pediatric oncology. The most prominent example is ALK, which was initially identified in anaplastic large cell lymphoma and later found in a significant share of neuroblastoma patients and in high frequency in lung cancer with activating mutation [43, 44]. Although the specific kinase inhibitor crizotinib was not very effective in the treatment of patients with high-risk neuroblastoma or relapse, newly developed ALK inhibitors such as ceritinib or lorlatinib may be more effective in this malignancy and are now being evaluated in clinical trials [29, 45, 46].

Analysis of gene-expression profiles can lead to the identification of patterns, which can then be targeted with different tyrosine kinases. Philadelphia-like ALL is exemplary here. The driver translocation t(9;22) of Philadelphia (Ph) ALL, results in the overexpression of the ABL oncogene, activating other tyrosine kinases itself. This spectrum of activation of different tyrosine kinases can also be found in samples of ALL missing the Ph translocation. This Philadelphia-like leukemia has an unfavorable prognosis, similar to Ph positive ALL, and can at least be co-treated successfully with tyrosine kinases as well [47].

There is a broad spectrum of diseases in pediatric oncology, where tyrosine kinase inhibitors can be used. Lesions thereof are being found in a minority of patients with a variety of cancers. Therefore, their use is only feasible and appropriate in the context of trials after sequencing has been performed. This concept of therapeutic personalization has lead to novel designs of clinical studies such as basket (same target in different entities) and umbrella (different targets in same entities) studies. Selected examples are shown in Table 1.

Apart from tyrosine kinase inhibitors, there is interest in other therapeutic strategies that aim to influence cell survival in general or target the “motor” independent of possible mutations [48‐51]. Amongst those are strategies that target proteins of the BCL (B-cell lymphoma)-2 family, which inhibits programmed cell death. Preclinical data suggests that high-risk patients in ALL or neuroblastoma could benefit from treatment with BCL-2 inhibitors [52, 53].

Here, too, findings in adult cancers were pioneering. Chronic lymphatic leukemia, defined by differentiation of B-lymphocytes and virtually untreatable through chemotherapy, shows an excellent and long-lasting response to the BCL-2 inhibitor venetoclax, which has recently been approved for clinical use [54].

Detailed genomic analysis has led to new definitions of different tumor entities, which were previously thought to be uniform. In medulloblastoma, there are four clearly distinguishable, molecularly defined subgroups with different genetic alterations that result in deregulated signal transduction, i.e., WNT (wingless), SHH (sonic hedgehog) (both named for altered signal transduction pathways), group, 3 and group 4. These subgroups are prognostically relevant, with WNT having the best and group 3 having the worst prognosis. Their molecular profiles can provide possible targets for approaches in precision medicine. Genome sequencing in the SHH subgroup, e.g., can predict whether a tumor is responsive to inhibition of the Smoothened (SMO) protein [55, 56].

The analysis of tumor genomes led to substantial insights into cancer development. Genomic analyses can provide biomarkers and identify novel targets for targeted therapies. Nevertheless, there are limitations: while there are bona fide examples for significant improvement of survival due to implementation of targeted therapies in adult cancers (such as treatment of EGFR-mutated lung cancer), not all hopes have been fulfilled due to primary or secondary resistance [57]. Furthermore, there is an intrinsic problem within the tumor itself. In many cases, tumors show extensive genetic heterogeneity, e.g., structural heterogeneity in osteosarcoma due to mutations in DNA repair [58]. Genetic heterogeneity can be found in solid tumors as well as in leukemia. It implies that different cells have different genetic alterations. It thus can be assumed that several clones exist at diagnosis. Relapse can arise by evolution from a preexistent subdominant clone resistant to therapy; these cells may not be detectable initially [59‐61]. Targeted therapy may thus have to aim at moving targets. This suggests a combination of therapies addressing different structures and signaling pathways. These results also suggest that the ability of the immune system to control and eliminate tumor cells has to be employed more often, if necessary, in combination.

In summary, the molecular analysis of tumors and leukemia in childhood and adolescence has made groundbreaking progress in our understanding of cancer. Furthermore, possible targets for specific therapies in the context of precision medicine have been identified. These therapeutic approaches may prove their efficacy in clinical trials and reduce the grave side effects of conventional cytotoxic therapies.

How long does it take from a scientific breakthrough in basic research to clinical application? 20, 50, or 100 years? All answers are correct for immunotherapy of cancer depending on which groundbreaking discovery you want to take into account [62‐71]. History reveals that translational research may reduce the latency period. Even in December 2013, when Science magazine picked cancer immunotherapy as the breakthrough of the year, there were still serious doubts amongst the jurors about whether this breakthrough would lead to a sustainable change in clinical practice. Today, immunotherapy is becoming the fifth modality in cancer therapy (next to surgery, radiation, chemotherapy, and targeted therapy).

The adaptive (or specific) immune system has two evolutionary related effector mechanisms: humoral and cellular immunity. Antibodies are the effectors of humoral immunity. They are the older extant within the evolution of the adaptive immune system. Antibodies are produced by B-lymphocytes and bind to molecules on the surface of target cells; thus, the repertoire of antibodies is limited to those target molecules that occur on the outer cell membrane of blood cells or cellular organisms circulating in the blood, i.e., bacteria. Effectors of humoral cytotoxicity are myeloid cells of the innate immune system (i.e., phagocytes, antibody dependent cellular cytotoxicity, ADCC) or the complement system (complement dependent cytotoxicity, CDC). Antibodies developed earlier than jawed vertebrates in evolution and the inborn immune system (innate, non-specific or natural immunity) is evolutionarily older than the adaptive immune system. It can already be found in plants.

Apart from their special teeth, carnivores were the first jawed vertebrates (gnathostomata) to develop a cellular adaptive immune system to reject the cells of their prey including incorporated pathogens and defend themselves against hostile takeover by their victims [72]. Effectors of cellular adaptive immunity system are T-lymphocytes. Through T-cell receptors (TCRs), they recognize endogenous and exogenous protein fragments (peptides) being presented by the major histocompatibility complex (MHC, in humans: human leukocyte antigens, HLA). In contrast to bacteria, viruses require host cells for replication and thus are primarily controlled by T-cells in contrast to B cells primarily controlling bacteria. However, the evolution of immunity, i.e., immunologic memory, requires interaction between the innate and adaptive immune system as well as B- and T-cells in the latter. Paradigmatically, every universal peptide can be recognized by the cellular adaptive immune system. Thus, the TCR-repertoire has been termed unlimited. However, the relation between the possibility of recombination of the TCR and the number of universally possible peptides implies an imperative TCR promiscuity: 10¹¹ human TCRs have to match with 10²⁰ peptides [73] (Fig. 3).

To execute the cytotoxic function, T-cells perforate the cellular membrane of the target cell with perforin and instill the cytotoxic protein granzyme B into the cytosol of the target cell.

More than 100 years ago (clearly before the advent of radio- and chemotherapy), the American surgeon William Coley treated sarcomas, amongst them Ewing sarcomas, successfully by inoculation with Coley’s toxin, a mixture of attenuated streptococci and serratiae. This unspecific immunotherapy is based on a stimulation of the inborn immunity resulting in an inflammatory reaction that can elicit an anti-tumor effect.

The high frequency of GC in mycobacterial DNA acts as a signaling pattern eliciting an innate immune response with consecutive T-cell stimulation. About 50 years ago, Mathé developed the inoculation with bacille Calmette-Guérin (BCG), which is still approved today in the treatment of bladder cancer. A newer innovation based on the immune stimulatory effect of mycobacteria is the macrophage-activating drug mifamurtide (muramyltripeptide, Mepact®). Mifamurtide is a synthetic analog of muramyldipeptide, an immunogenic component of the mycobacterial pathway. The immune stimulatory effect of mifamurtide is mediated via the binding of NOD2 (nucleotide-binding oligomerization domain protein)-receptors on monocytes, macrophages, and dendritic cells. Mifamurtide was approved in the EU in 2009, about 30 years after it was developed, as an orphan drug for the treatment of osteosarcoma. Apart from myeloid-derived monocytes and macrophages, lymphoid-derived natural killer cells (NK) are the cellular effectors of innate immunity. NK cells can elicit antitumor effects, especially in acute myeloid leukemia and solid tumors in adults.

Table 2 gives an overview of progress in precision medicine in immuno- and cell-based therapy in pediatric oncology.

The centennial success of pediatric oncology was based on the multidisciplinary approach involving less mutilating surgery in a neo-adjuvant setting as well the cytotoxic modalities of mutagenic cell toxins (e.g., war-agent derivatives) and ionizing radiation with ensuing long-term toxicity in cancer survivors. Disruptive high-throughput technologies may provide an urgently needed paradigm shift here. The current concept of randomized trials may not fully appreciate the heterogeneity of increasingly subdivided entities and is being replaced by individualized therapies according to genomic and other high throughput analyses. This individualization implies the risk that the efficacy (and even more superiority) of these novel therapies may be difficult to prove when only a few patients are treated with a specific treatment regimen. These challenges have to be addressed by novel study concepts, including adaptive design, basket and umbrella trials, establishing surrogate endpoints (e.g., biomarkers) as well as multi-modal high throughput molecular analyses of individual patients. This individualization of therapy will put the individual patient into the focus of research.

The imminent paradigm shift is based on the assumption that individualized molecular analysis will provide biomarkers for targeted therapies that eliminate malignant but spare normal cells. This assumption has not been proven yet for tumor stem cells. An alternative assumption implies that development of cancer in children is not characterized by an accumulation of oncogenic events, but by a selection advantage of any genetic event favoring dedifferentiation and the reversion to the embryonic default mode. Along with this assumption, it may not be the genetic alterations that are carcinogenic, but a tumor-microenvironment may provide epigenetic and metabolic reprogramming [96]. Reprograming T-cells with chimeric receptors to manipulate selectively that tumor microenvironment may open here new horizons for cancer immunotherapy [97].