Simultaneous Brg1 Knockout and MYCN Overexpression in Cerebellar Granule Neuron Precursors Is Insufficient to Drive Tumor Formation but Temporarily Enhances their Proliferation and Delays their Migration

Medulloblastoma (MB) is the most common malignant brain tumor in childhood. According to the World Health Organization (WHO) classification of central nervous system (CNS) tumors, this embryonal tumor is divided into a wingless (WNT)-activated, Sonic hedgehog (SHH)-activated, and non-WNT/non-SHH entity. The latter is poorly defined but frequently carries mutations in Brahma-related gene 1 (BRG1) or amplifications of MYCN. Here, we investigated whether a combination of a Brg1 knockout and an overexpression of MYCN in cerebellar granule neuron precursors or multipotent neural stem cells is sufficient to drive brain tumor formation in mice. To this end, we generated Math1-creERT2::Brg1fl/fl::lslMYCN and hGFAP-cre::Brg1fl/fl::lslMYCN mice, respectively. We did not observe brain tumor formation in any of these models. hGFAP-cre::Brg1fl/fl::lslMYCN mice revealed severe CNS abnormalities with short survival, similar to the situation with a sole loss of Brg1, as we previously described. Investigation of Math1-creERT2::Brg1fl/fl::lslMYCN mice with a tamoxifen induction at postnatal day 3 revealed a regular survival but significant increase in cerebellar granule neuron precursor proliferation, followed by a delayed inward migration of these cells. This is in stark contrast to the hypoplastic cerebellum that we previously observed after embryonic deletion of Brg1 in Math1 positive cerebellar granule neurons. Our results indicate a time-specific function of Brg1 in cerebellar granule neuron precursors. Yet, the exact temporal and spatial origin of non-WNT/non-SHH MB remains unclear.


Introduction
Medulloblastoma (MB) is the most common malignant brain tumor in children [1]. This embryonal tumor arises in the posterior fossa and can be divided into at least three molecular subgroups: wingless (WNT)-activated MB, Sonic hedgehog (SHH)-activated MB, and non-WNT/non-SHH MB [2]. WNT and SHH MBs are characterized by activating mutations in the respective pathways, but driver mutations in non-WNT/non-SHH, which can be further segregated into Group 3 and Group 4 MBs, are less well understood [3]. A number of large-scale sequencing studies revealed that missense mutations in the epigenetic modifier Brahma-related gene 1 (BRG1, SMARCA4, BAF190A) are found in 4.3-8.8% of all MBs, but especially in WNT and Group 3 MBs [4][5][6]. In the study performed by Jones et al., BRG1 was the most common mutated gene in Group 3 MB. BRG1 encodes one of the two mutually exclusive ATPase subunits of the SWItch/sucrose nonfermenting (SWI/SNF) chromatin remodeling complex, which regulates gene expression by increasing nucleosome mobility [7][8][9]. It has been described as a tumor suppressor in different cancer entities, such as rhabdoid tumors, small cell carcinoma of the ovary, hypercalcemic type (SCCOHT), and lung cancer [10][11][12][13][14]. Still, a Brg1 knockout in various Dörthe Holdhof and Ji Hoon On contributed equally to this work. different neural stem cell (NSC) populations in mice was not sufficient to drive tumor development but caused hypoplasia of diverse brain regions [15][16][17]. Amplifications of the protooncogene MYCN have been identified in several cancer entities, including tumors of the central nervous system (CNS) such as high-grade gliomas, spinal ependymoma and MB [18][19][20][21]. In MB, they are found in SHH, Group 3 and Group 4 MBs and are associated with a poor outcome [4-6, 19, 22]. MYCN is a transcription factor of the MYC family and is important for cell growth, apoptosis, tumor cell metabolism and normal cerebellar development [21,23,24]. Swartling et al. demonstrated that overexpression of MYCN in Glutamate transporter 1 (Glt1) positive cells gives rise to murine MB [25]. However, these tumors occurred rather late in development and accumulated sporadic p53 mutations in addition to the MYCN alteration [26]. Furthermore, due to the broad expression of the promoter and the late tumor detection, the exact anatomical origin of these tumors remains obscure. The same group recently reported that ectopic expression of MYCN in human induced pluripotent stem cell-derived neuroepithelial stem (NES) cells increases proliferation in vitro. Orthotopic transplantation of these NES cells into mice results in tumor formation resembling human SHH MB, again without precise information about the tumor origin [27]. In hGFAP expressing cells, MYCN amplification did not result in MB formation, possibly due to the lack of a second hit [28]. In this study, we were interested whether Brg1 knockout and MYCN overexpression might have synergistic effects in the development of tumors in the posterior fossa. Therefore, we generated Math1-creER T 2 ::Brg1 f l / fl ::lslMYCN fl/fl (Mert::B fl/fl ::N fl/fl ) and hGFAP-cre::Brg1 fl/ fl ::lslMYCN fl/fl (hG::B fl/fl ::N fl/fl ) mice to investigate the combination in postnatal granule neuron precursor cells (GNPCs) and multipotent NSCs, respectively [29,30]. Mice carrying the inducible creER T2 transgene under the control of the Math1 promoter received tamoxifen at postnatal day 3 (P3) to induce the cre recombinase. In hG::B fl/fl ::N fl/fl mice, the cre recombinase was constitutively active from embryonal day 13.5 (E13.5) onwards [30]. We observed that the loss of Brg1 in combination with an additional overexpression of MYCN in Math1 expressing cells resulted in changes regarding cell survival parameters in the external granular layer (EGL) of the cerebellum. Furthermore, GNPCs showed delayed inward migration. However, the adult cerebellum appeared normal with regular organization into inner granular layer (IGL), Purkinje cell layer (PCL) and molecular layer (ML) and without any signs of tumor development. Amplification of MYCN in addition to the loss of Brg1 in hGFAP expressing cells did not rescue the phenotype caused by the deprivation of Brg1 alone. As we published earlier, Brg1 deficiency in hGFAP positive cells causes a hypoplastic cerebellum [17]. Likewise, hG::B fl/fl ::N fl/wt , and hG::B fl/fl ::N fl/ fl mice presented with an underdeveloped cerebellum without organization into lobuli and with absence of the characteristic cerebellar layering. They died after about 2 weeks without any signs of tumor formation. Hence, we conclude that MYCN amplification and Brg1 deficiency disturb cerebellar development but are not sufficient to drive brain tumors originating from hGFAP or Math1 positive precursor cells.

Image Quantifications and Statistical Analysis
In order to quantify the proportion of cells positive for a specific marker, we used IHC stained sagittal sections and counted their numbers in relation to the total number of cells per field using the ImageJ software (Wayne Rasband, National Institute of Health, USA). For marker expression in the EGL, we quantified the cells in the EGL between lobuli V and VI. To analyze the proportion of Pax6 positive cells in the ML, we counted the cells in lobulus VI. Marker expression was quantified for at least three animals per genotype by a blinded investigator. All statistical analyses were performed using Prism Software Version 7 (GraphPad Software, Inc., San Diego, USA). For quantifications of marker expression in the EGL and the ML, a Tukey's multiple comparisons test and for survival analyses, a log-rank Mantel-Cox test was performed.

A Postnatal Loss of Brg1 With or Without Additional Amplification of MYCN in Math1 Expressing Cells Causes an Increase in Proliferation
A prenatal loss of Brg1 in Math1 positive cells results in cerebellar hypoplasia [16]. Therefore, we chose the tamoxifen-inducible creER T2 recombinase to generate mice with a postnatal Brg1 loss either alone or with additional overexpression of the human MYCN in Math1 expressing cells. We injected tamoxifen at P3 to activate cre-mediated recombination. In a first approach, we were interested in the shortterm effects and sacrificed the animals 5 days after tamoxifen injections at P8. At this time, proliferation of GNPCs peaks during normal murine cerebellar development [34,35]. H&E stains revealed no morphological alterations at P8 as animals of all genotypes had normal cerebella with regular lobuli and distinct layers ( Fig. 1a, b, h, i, o, p, v, w, ac, ad). We further investigated the extent of Brg1 loss and MYCN overexpression by IHC. In controls and in Mert::B fl/wt animals, all cells were positive for Brg1 and MYCN was expressed in the outer EGL ( Fig. 1c, d, j, k). The EGL of the remaining three genotypes consisted of Brg1 negative cells in the outer EGL and Brg1 positive cells in the inner EGL (Fig. 1q, x, ae). This result was expected, since we induced Brg1 loss by tamoxifen treatment not before P3. The number of MYCN expressing cells was significantly increased in the presence of the lslMYCN transgenes (71% and 89% in Mert::B fl/fl ::N fl/wt and Mert::B fl/fl ::N fl/fl , respectively) compared to controls (33%), Mert::B fl/wt (31%) and Mert::B fl/fl (38%) mice (Fig. 1y, af, aj).
Next, we investigated the influence of Brg1 loss and MYCN overexpression on cell survival by examining apoptosis and proliferation marker expression. We stained for cl.
Casp3 as an indicator for apoptosis induction (Fig. 1e, l, s, z, ag, ak). Quantification of cl. Casp3 stainings revealed no significant differences in the number of apoptotic cells in the EGL of control mice (0.5%) and mice with both a hetero-(0.47%) and a homozygous loss of Brg1 (0.76%). However, in Mert::B fl/fl ::N fl/fl mice, but not in Mert::B fl/fl ::N fl/wt mice, the additional overexpression of MYCN caused a significant increase in apoptosis in the EGL as 2.24% and 1.36% of cells were cl. Casp3 positive, respectively (Fig. 1ak). In order to investigate proliferation, we pulsed the animals 2 h prior to sacrifice with BrdU, a thymidine analog that is incorporated into newly synthesized DNA during the S phase of the cell cycle. We stained for BrdU and additionally for pHH3 to examine two independent proliferation markers ( Fig. 1f, g, m, n, t, u, aa, ab, ah, ai). There were no significant differences in the number of BrdU positive cells in the EGL of control (27.0%), Mert::B fl/wt (23.6%) and Mert::B fl/fl mice (32.7%), even though there was a trend towards more proliferating cells in the latter (Fig. 1al). However, in case of a simultaneous MYCN amplification, there were significantly more BrdU positive cells in the EGL. In Mert::B fl/fl ::N fl/wt mice, 40.2% of cells and in Mert::B fl/fl ::N fl/fl mice, 39.5% of cells had incorporated BrdU during S phase (Fig. 1al). Likewise, the percentage of pHH3 expressing cells was insignificantly different in case of a Brg1 knockout without MYCN amplification (Fig.  1am). In the EGL of controls, Mert::B fl/wt mice and Mert::B fl/fl mice, 7.86%, 9.42%, and 10.8% were pHH3 positive, respectively. Again, these results indicate that a Brg1 deficiency caused a trend towards more proliferating cells. Amplification of MYCN caused a significant rise in the number of pHH3 expressing cells in the EGL of Mert::B fl/fl ::N fl/fl mice (14.5%), but not of Mert::B fl/fl ::N fl/wt mice (11.2%).
Taken together, the combination of a Brg1 deficiency with MYCN overexpression significantly alters the expression of cell survival parameters in the EGL at P8.

Brg1 Deficiency and MYCN Amplification Decreases Inward Migration of Granule Neurons
In mice, cerebellar cortical development is finalized at an age of 3 weeks resulting in three distinct layers: ML, PCL and IGL [36]. In order to analyze how the postnatal knockout of Brg1 with or without simultaneous overexpression of MYCN influenced the ontogenesis of the cerebellum, we sacrificed the mice at P21 and investigated the morphology by H&E stains (Fig. 2a, b, e, f, i, j, m, n, q, r). In Mert::B fl/fl mice, it appeared as if more cells were present in the ML compared to control and Mert::B fl/wt mice. This effect seemed to be even more pronounced in Mert::B fl/fl ::N fl/wt mice and Mert::B fl/fl ::N fl/fl mice. To validate, whether the supernumerary cells in the ML were Brg1 deficient, we stained for Brg1 (Fig. 2c, g, k, o, s). In all mice not carrying a homozygously floxed Brg1 allele, Brg1 was expressed in all cells. In all three mutants, Brg1 negative cells were not only observed in the IGL, but also in the ML. This suggested that Brg1 loss resulted in decelerated migration of GNPCs from the EGL to the IGL. In order to verify that these cells were GNPCs, we stained for Pax6 and quantified the proportion of Pax6 expressing cells in the ML (Fig. 2d, h (Fig. 2u-x). Brg1 negative cells were found in both, the ML and the IGL, and only a very small fraction of Pax6 expressing cells was present in the ML, similar to the expression profile of control P21 cerebella. We therefore conclude that GNPC inward migration is delayed after loss of Brg1 and overexpression of MYCN but is completed at the age of 6 months. In order to exclude tumor growth in a small proportion of such animals, we observed a total of 14 Mert::B fl/fl ::N fl/wt and 12 Mert::B fl/fl ::N fl/fl animals over a period of 6 months ( Supplementary Fig. 1a). Three and two animals of these cohorts died, respectively, without any preceding symptoms. Therefore, we were not able to analyze their brains or bodies. All other animals underwent full autopsy after 6 months, but we did not detect any tumor growth or other abnormalities. Although we cannot report on the autopsy of 5 out of 25 animals, we assume that a simultaneous loss of Brg1 and overexpression of MYCN in Math1 positive GNPCs is not sufficient to drive tumor growth.

Amplification of MYCN in Multipotent NSCs Does Not Rescue the Effects Caused by the Loss of Brg1
In order to examine the combination of Brg1 deficiency and MYCN amplification in a spatially and temporally different setting, we generated hG::B::N mice. In contrast to the induced cre recombinase activity in postnatal GNCPs in Mert::B::N animals, hGFAP drives the constitutive expression of the cre recombinase in multipotent NSCs from E13.5 onwards. Consequently, Brg1 loss and MYCN overexpression can be observed in almost all cerebellar cells, except for Purkinje cells and choroid plexus epithelium [30]. We previously reported that multipotent NSCs marked by hGFAP expression depend on the proper expression of Brg1 to give rise to all major brain regions [17]. For instance, the cerebella of hGFAP-cre::Brg1 fl/fl mice are hypoplastic without foliation or lamination. In the present study, we investigated, whether an additional MYCN amplification was able to rescue this phenotype or would even give rise to brain tumors. For this purpose, we generated hG::B fl/wt ::N fl/wt , hG::B fl/wt ::N fl/fl , hG::B flfl ::N fl/wt and hG::B flfl ::N fl/fl mice. In case of a heterozygous Brg1 deficiency, the cerebella at P8 appeared normal in H&E stains compared to controls (Fig. 3a, b, f, g, k, l). As expected, Brg1 was detectable in all cells of these mice (Fig. 3c, h, m), whereas the number of MYCN positive cells seemed increased in the EGL of hG::B fl/wt ::N fl/wt and hG::B fl/wt ::N fl/fl mice (Fig. 3d, i, n)In contrast, the cerebella of hG::B flfl . ::N fl/wt and hG::B flfl ::N fl/fl mice were severely underdeveloped and resembled those of hGFAP-cre:.Brg1 fl/fl mice. Neither lobules nor distinct layers were detectable in H&E stains (Fig. 3p, q, u, v). Like in hGFAP-cre::Brg1 fl/fl mice, Brg1 expression was lost in a fraction of cells distributed across the entire cerebellum of hG::B flfl ::N fl/wt and hG::B flfl ::N fl/fl mice (Fig. 3r, w). As the hGFAP promoter is expressed in the majority of cerebellar cells or their ancestors [30], Brg1 deficiency was likely to be present in most cell types. However, only a small proportion of cells were positive for MYCN (Fig. 3s, x). This indicates either that the recombination of two floxed transgenes was not efficient or that especially those cells with enhanced MYCN expression were depleted from the developing brain until P8. Next, we examined, whether the absence of distinct layers in hG::B flfl ::N fl/wt and hG::B flfl ::N fl/fl cerebella was caused by impaired neuronal migration due to altered Bergmann glia cell morphology. Proper neuronal migration is highly dependent on Bergmann glia cells [37], and hGFAP-mediated cre expression is present in these cells [30]. We employed S100 as a Bergmann glia marker [38] that stains the entire cells including their radial processes. In controls, hG::B fl/wt ::N fl/wt and hG::B fl/wt ::N fl/fl mice, Bergmann glia cells were located in the PCL and their processes extended to the pial surface (Fig. 3e, j, o). In hG::B f l f l ::N f l / w t and hG::B flfl ::N fl/fl cerebella, there were almost no S100 positive cells detectable and those that were present did not have any radial processes (Fig. 3t, y).
Finally, we investigated the mice for the development of disease-related symptoms for half a year ( Supplementary  Fig. 1b). Fifteen out of 33 animals with a homozygous loss of Brg1 had to be sacrificed in the first 2 weeks of life, indicating that the overexpression of MYCN did not compensate for the loss of Brg1. In the few animals that survived into adulthood, we did not observe any Brg1 deficient cells in their brains (data not shown). This suggests that in these mice, the simultaneous recombination of the two floxed transgenes (Brg1 and lslMYCN) was not efficient, causing long-term survival. Of note, none of these mice developed a brain tumor. However, some hG::B fl/wt ::N fl/wt , hG::B fl/wt ::N fl/fl and hG::B fl/fl ::N fl/wt mice died due to neuroendocrine tumors of the pancreas and the pituitary gland as already reported for hGFAP-cre::lslMYCN mice (data not shown) [28]. Hence, the development of these tumors was likely caused by the amplification of MYCN but not by the Brg1 deficiency.
In summary, we did not observe any cooperative effects of Brg1 loss and MYCN amplification in these mice, as hG::B fl/wt ::N fl/wt and hG::B fl/wt ::N fl/fl mice developed similar phenotypes as hGFAP-cre::lslMYCN mice. Likewise, hG::B flfl ::N fl/wt and hG::B flfl ::N fl/fl mice developed morphological alterations in the brain, which resembled the abnormalities observed in hGFAPcre:.Brg1 fl/fl mice.

Discussion
Mutations in BRG1 and amplifications of MYCN have frequently been reported in MBs, especially those of the non-WNT/non-SHH subgroup. This subgroup comprises two thirds of all MBs, but the molecular mechanisms driving oncogenesis are insufficiently understood [4-6, 19, 39]. Therefore, the aim of this study was to elucidate, whether the combination of Brg1 loss and MYCN overexpression in postnatal GNPCs or multipotent NSCs in mice could model the development of human non-WNT/non-SHH MB. To this end, we used Mert::B::N and hG::B::N mice, and examined histological alterations and survival of these animals.
First, we studied the short-term effects of a postnatal Brg1 loss and MYCN amplification on cell survival parameters of the CGNPs in the EGL of Mert::B::N mice. At P8, MYCN overexpression in a Brg1 deprived EGL caused a significant increase in both, proliferating and apoptotic cells. Normal cerebellar development is a highly orchestrated sequence of proliferation, differentiation, migration and programmed cell death. The latter is important to eliminate supernumerary cells from the developing brain (reviewed by [40]). Hence, the increase in cl. Casp3 expressing cells indicating an increase of apoptotic cells in the EGL might just be a compensatory mechanism for the enhancement of proliferation caused by Brg1 loss and MYCN overexpression. MYCN is known to be essential for SHH-mediated CGNP proliferation and is amplified and upregulated in human and murine MB, respectively [21,24,41,42]. Previous results already suggested that the enhancement in proliferation caused by MYCN amplification alone does not result in MB formation from CGNPs [28]. Even though the postnatal loss of Brg1 alone appeared to show a trend towards enhanced proliferation, the combination of MYCN overexpression and Brg1 loss did not synergize and did still not result in tumor formation. Nonetheless, we cannot completely preclude that knockdown of Brg1 and overexpression of MYCN might not be fully parallelized events as it might take longer/shorter to accumulate MYCN protein than to lose the Brg1 protein.
The finding of a slightly increased (although not significant) EGL proliferation in Mert::B fl/fl mice after application of tamoxifen at P3 is still noteworthy, since it contradicts previous results from (non-inducible) Math1-cre::Brg1 fl/fl mice, in which the ATPase is lost at around E10.5. The latter suffer from a severely hypoplastic cerebellum caused by decreased proliferation of GNPCs [16]. Similarly, Zhan et al. reported that loss of Brg1 at P0 in Nestin-creER T2 ::Brg1 fl/fl mice resulted in smaller cerebella [43]. Together, in combination with these published data, our here presented results highlight the time-specific role of Brg1 in the developing cerebellum, a phenomenon that has also been described for other major players in cerebellar development [44,45].
In addition to the changes in proliferation and apoptosis at P8, we recognized that the induced Brg1 deficiency in combination with MYCN amplification at P3 resulted in the accumulation of cells in the ML at P21. During murine cerebellar development, CGNPs in the EGL first stop proliferating, then start to differentiate and finally migrate towards the IGL, where they start to form synaptic connections [46]. Quantification of Pax6 positive cells revealed that Brg1 loss by itself or in combination with MYCN amplification caused a significant increase in the proportion of granule cells in the ML. This indicates that the migratory behavior of granule neurons was impaired in Mert::B fl/fl , Mert::B fl/fl ::N fl/wt and Mert::B fl/fl ::N fl/fl mice. Of note, many of the cells in the ML were negative for Brg1 in IHC, supporting the hypothesis that the genetic alterations in our mutant mice were directly responsible for the observed phenotype. Likewise, Brg1 deprivation in cerebellar explants from hGFAP-cre::Brg1 fl/fl mice lead to decreased migration of late-migrating cells in vitro [17]. Additionally, in hG::B flfl ::N fl/wt and hG::B flfl ::N fl/fl mice, the number and morphology of Bergmann glia was severely diminished, indicating that the observed lack of layering in these mice, might be partially explained by disturbed migration as well. Our results suggest that the addition of MYCN overexpression enhanced the migration deficit caused by the Brg1 deficiency. MYCN is essential for normal cerebellar development as indicated by different mouse models examining MYCN knockout [23,47]. However, to our knowledge the need for balanced MYCN levels for granule cell migration has not been described before. In the adult mice, i.e. 6 months or older, there were no misplaced Pax6 positive cells left in the ML, indicating that granule cell migration was delayed but not permanently inhibited in our model.
A fraction of hG::B::N mice developed neuroendocrine tumors of the pancreas and the pituitary gland as previously published for hGFAP-cre::lslMYCN mice [28]. Consequently, in these mice, the amplification of MYCN determined the phenotype, whereas the heterozygous Brg1 loss (or the incomplete recombination in mice with a homozygous Brg1 loss) had no obvious impact. Vice versa, hG::B flfl ::N fl/wt and hG::B flfl ::N fl/fl mice presented with symptoms resembling those of hGFAP-cre::Brg1 fl/fl mice, including early postnatal death, hydrocephalus and hypoplasia of the cerebrum and cerebellum [17]. In these mice, the effect of the homozygous Brg1 deficiency was dominant and the additional overexpression of MYCN did not seem to have an impact. However, as we did not detect MYCN protein expression by IHC in the majority of cerebellar cells at P8, we cannot rule out that the recombination efficiency of the loxP sites of the lslMYCN transgene was decreased in case of the additional presence of a homozygously floxed Brg1 gene. Another more likely explanation for the lack of MYCN expressing cells might be that most of the cells hit by the loss of Brg1 and MYCN overexpression have not at all developed or have died by P8. Most importantly, Brg1 deficiency in combination with overexpression of MYCN in hGFAP positive NSCs does not result in brain tumor formation.

Conclusion
In this study, we demonstrate that the combination of a Brg1 deficiency with MYCN amplification in mice is not sufficient to drive tumor formation, neither in postnatal Math1 expressing cells nor in hGFAP positive multipotent NSCs. Consequently, either these cells do not represent cells of origin for non-WNT/non-SHH MBs or the combination of Brg1 loss and MYCN amplification does not provide the genetic basis for tumorigenesis. Still, the animals presented with minor developmental disturbances. Furthermore, in Math1 positive cells, the combination of the two genetic events seems to have additive effects regarding survival and migration of granule neurons. In contrast, the homozygous knockout of Brg1 defined the phenotype in hG::B flfl ::N fl/wt and hG::B flfl ::N fl/fl mice, whereas the addition of MYCN amplification had no obvious additional effect.
Acknowledgments We thank Dr. J. Schulte (Charité Berlin, Germany) for providing lslMYCN fl/fl mice. We further thank Anne Reichstein, Jacqueline Kolanski, Vanessa Thaden, and Kristin Hartmann for expert technical assistance. Finally, we acknowledge the support of the Small animal models core facility of the Heinrich Pette Institute.
Author Contribution All authors contributed to the study conception and design. Material preparation, data collection, and analysis were performed by Dörthe Holdhof and Ji Hoon On. The first draft of the manuscript was written by Dörthe Holdhof. All authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.
Funding Open Access funding enabled and organized by Projekt DEAL. This work received funding from the Deutsche Krebshilfe, and U.S. was additionally supported by the Fördergemeinschaft Kinderkrebszentrum Hamburg. JHO was supported by a Juli Harnack fellowship for medical students from the Fördergemeinschaft Kinderkrebszentrum Hamburg.

Compliance with Ethical Standards
Conflict of Interest The authors declare that they have no conflict of interest.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.