Clinicopathological features and prognostic value of SOX11 in childhood acute lymphoblastic leukemia

Acute lymphoblastic leukemia is marked by aberrant transcriptional features that alter cell differentiation, self-renewal, and proliferative features. We sought to identify the transcription factors exhibiting altered and subtype-specific expression patterns in B-ALL and report here that SOX11, a developmental and neuronal transcription factor, is aberrantly expressed in the ETV6-RUNX1 and TCF3-PBX1 subtypes of acute B-cell leukemias. We show that a high expression of SOX11 leads to alterations of gene expression that are typically associated with cell adhesion, migration, and differentiation. A high expression is associated with DNA hypomethylation at the SOX11 locus and a favorable outcome. The results indicate that SOX11 expression marks a group of patients with good outcomes and thereby prompts further study of its use as a biomarker.

JURKAT, MOLT-16, P12-ICHIKAWA, HPB-ALL, and CCRF-CEM were cultured in RPMI Medium 1640 (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) with 2 mM L-glut, 100 U penicillin, 100 µg/ml streptomycin with 10% FBS (Gibco), and MOLT-4, PEER, and MHH-CALL3 with 20% FBS (Gibco) at 37 °C in 5% CO 2 . An inducible E/R fusion in the NALM-6 cell line and a knockdown of E/R by a short hairpin RNA (shRNA) in the REH cell line have been previously described 37 . E/R expression was induced with 500 ng/ml doxycycline (Clontech). E/R expression changes were confirmed with RT-qPCR, with fusion gene-specific primers (Table S2). Mycoplasma tests were done regularly for the cell lines, and Eurofins Genomics (Ebersberg, Germany) services were used to authenticate the cell lines by STR genotyping. All cell lines used in this study were purchased from the Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany).
For the methyltransferase inhibition experiments, the cultured cells were treated for 72 h with decitabine, 5-Aza-2′-deoxycytidine solved in DMSO (A3656, Sigma-Aldrich, St. Louis, MO, USA) at 0, 0.1, and 1 µM concentrations. The media were changed at a 24 h interval to compensate for decitabine instability under cell culture conditions. After the treatment cells were collected, RNA was extracted for RT-qPCR analyses. Corticosteroid and chemotherapy treatments were conducted with the indicated concentration ranges, and cell viabilities were measured after either 72 (697 and RCH-ACV) or 96 h (REH). The corticosteroids included prednisolone (P6004, Sigma-Aldrich) and dexamethasone (D8893, Sigma-Aldrich), and the chemotherapy agents used were asparaginase (A3809, Sigma-Aldrich) and vincristine (V8879, Sigma-Aldrich). The applied concentrations for each cell line are indicated in Table S3.
Quantitative real-time PCR. Total RNA was extracted using the PureLink ™ RNA Mini Kit, and the On-Column PureLink ® DNase Treatment Protocol was used for DNA removal (Ambion ® by Life Technologies and Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA); 100-500 ng of the extracted RNA was used as a starting material for cDNA synthesis, which was performed with iScript (Bio-Rad, Hercules, CA, USA). RT-qPCR reactions were conducted according to the manufacturer's instructions with SsoFast EvaGreen ® Supermix (Bio-Rad). The following program was performed with the Bio-Rad CFX96 TM Real-Time System (Bio-Rad): initial denaturation at 96 °C for 30 s, 39 cycles of denaturation at 96 °C for 2 s, annealing at 60 °C for 5 s, and plate read. Independent experiments performed in triplicate were used as the starting material for the RT-qPCR measurements, and the relative 2 −ΔΔC T method was used for quantification 38 . The primer sequences used in RT-qPCR are listed in Table S2.
Western blot. Protein extraction was performed using M-PER reagent (Thermo Fisher Scientific) according to the manufacturer's instructions to lyse the cells, and protein concentrations were measured with DC Protein Assay (Bio-Rad); 15-20 µg of a protein sample was loaded into the precast 10% Mini-PROTEAN ® TGX Stain-Free TM Gels (Bio-Rad). After the electrophoresis run, Trans-Blot ® Turbo TM Pack (Bio-Rad) was used to transfer the proteins from the gel to the nitrocellulose membrane. Transfer was done with the Trans-Blot ® Turbo TM Transfer system according to the manufacturer's instructions (Bio-Rad). A prestained protein ladder, PageRuler Plus (#26619, Thermo Fischer Scientific), was used as a protein size marker. We utilized antibodies against SOX11 (1:1,000) (HPA000536, Sigma-Aldrich, Lot # BB107024) and Histone H3 (1:75,000) (ab4729, Abcam, Cambridge, UK, Lot # GR167613-1). Horseradish peroxidase conjugated anti-rabbit (1:2,000) (P0217, Lot # 00069121) was used as a secondary antibody (Agilent Technologies, Santa Clara, CA, USA). Chemiluminescence reaction by Amersham ECL reagent was detected with ChemiDoc TM XRS+using Image Lab TM Software (Bio-Rad).
Gene silencing with nucleofection. Knockdowns for suspension cells were performed using 4D-Nucleofector ™ (Lonza, Basel, Switzerland) for transfections. SOX11 knockdown was done using gene sequence-specific small interfering RNA (siRNA) (Sigma-Aldrich), and a non-specific siRNA was used as a control (Table S4). Before nucleofection, 20 µM stock solutions were diluted in Resuspension Buffer (SR30005, OriGene, Rockville, MD, USA) so that the final concentration per reaction was 300 nM. One million cells were used for each nucleofection reaction. Nucleofection reactions were conducted in proper solutions and programs according to the manufacturer's instructions in single nucleocuvettes (Table S5). Then, the cells were transferred to 12-well plates with prewarmed fresh media. The transfected cells were used in cell viability assays and RNA sequencing (RNA-seq). Western blot and RT-qPCR were used to assess knockdown levels.
RNA sequencing of cell lines. SOX11-specific and control siRNAs were used in the nucleofection for the 697, RCH-ACV, and REH cell lines to create SOX11 knockdown and control samples. After 48 h of nucleofection, three million cells per sample were collected for the total RNA extraction. RNA extraction was performed with the PureLink ™ RNA Mini Kit, and the On-Column PureLink ® DNase Treatment Protocol was used to avoid contamination by genomic DNA (Ambion ® by Life Technologies and Invitrogen). Protein samples were also collected, and the SOX11 knockdown level was verified by both RT-qPCR and Western blotting. Three independent biological replicates were collected for each transfected cell line, and each sample had 25 ng of RNA in 40 µl. Library preparation and RNA-seq (GSE123943) were performed in the Finnish Functional Genomics Centre (Turku, Finland). See more details in Supplementary Information.
The quality of the raw sequencing reads was ensured with FastQC (v0.10.1). Based on the FastQC results, reads were trimmed and their quality was filtered using the FASTX Toolkit (0.0.14). The reads were mapped to the human reference genome version hg19 using STAR aligner software (2.5.3a modified); reads aligning to more than two locations were discarded 39 . The alignment file was turned into tag directories, and read counts were calculated using the HOMER tool kit (v4.8).
Differential gene expression was analyzed using the quasi-likelihood F-test from edgeR, an R package 40 . Differentially expressed genes from all cell lines were filtered using the adjusted p-value < 0.05 (Benjamini-Hochberg method) as a cut-off, and the resulting gene lists from three different cell lines were compared by drawing a Venn diagram with an interactive Venn online tool (http://www.interactivenn.net/). Heatmaps, presenting all biological replicates, were drawn by using the z-scores of reads per kilobase of transcript, the per million mapped reads (RPKM) normalized count matrix, and the ComplexHeatmap R-package 41 . Gene set enrichment was analyzed by Gene Set Enrichment Analysis (GSEA) 3.0 software using logFC ranked lists of genes from differential gene expression analysis, and the results presented had an adjusted p-value < 0.02 (Benjamini-Hochberg method) 42,43 . Gene ontology (GO) term enrichment was studied with two approaches-with GSEA software using f-statistics ranked lists of genes and with the GOrilla online tool using unranked target and background lists 44,45 . The target lists were created with a threshold of adjusted p-value < 0.05 and logFC >0.5 or <−0.5.

RNA sequencing and methylation analysis of the patient samples. Previous transcriptome
sequencing of pediatric ALL cohort by Marincevic-Zuniga et al. 46 included 116 BCP-ALL cases, of which 115 cases were used to assess the SOX11 mRNA expression level in this study. DNA methylation data (GSE49031) were available for 112 of the 115 BCP-ALL cases 35 . See more details in Supplementary Information. Clinical data and the patient samples. The data on pediatric ALL cases below 18 years of age and treated in the Tampere University Hospital were retrieved from the clinical registry from years 1990 to 2017. Essential clinical information, such as age, leukocyte count, gender, relapse, death, central nervous system leukemia, immunophenotype based on flow cytometry, and clinical genetic information, was collected. The study obtained permission from the local ethical committee, and informed consent was sought, as needed (Pirkanmaa Hospital District Ethical Committee, R16054 and R13109). The use of old biopsy samples was approved by the National Supervisory Authority for Welfare and Heath (Valvira), and the samples were handled in accordance with relevant guidelines and regulations.

Cell viability and proliferation assays.
A total of 126 representative diagnostic formalin-fixed and paraffin-embedded (FFPE) decalcified bone marrow trephine biopsy samples were collected from the pathology department archives on the basis of the primary sample reports. For a proportion of cases, remission and relapse samples were also retrieved. Plastic-embedded and inadequate samples were excluded. The cases were classified based on the WHO 2017 Classification of Tumors and Haematopoietic and Lymphoid Tissues 47 .
Immunohistochemistry. Four micron-thick whole tissue sections were used for immunohistochemistry.
All cases were stained with anti-SOX11 antibody (clone MRQ-58, Cell Marque, Sigma-Aldrich, Lot # 1331005 A and 1430213 C) using Ventana Benchmark Classic at a dilution of 1:50. FFPE human MCL was used as a positive control material, and remission bone marrow samples were used as negative controls.
Staining intensity was graded, and cases with less than 20% of positivity of leukemic blast cell nuclei were interpreted as negative. Cases with immunohistochemical positivity ranging from 20% to 50% in the blast cell nuclei were graded positive, and cases with immunohistochemical nuclear positivity of over 50% were considered strongly positive. SOX11 expression was independently analyzed with a light microscope by two experienced pathologists without knowledge of the clinical data. Cases with discrepant scores were further analyzed by a third pathologist. www.nature.com/scientificreports www.nature.com/scientificreports/ acquired with the Beckman Coulter Navios cytometer (Beckman Coulter, Brea, CA, USA) using Red laser (638 nm) and 660/20 bandpass filter. Data analysis was conducted using Kaluza software (Beckman Coulter).
Statistical analysis. IBM SPSS Statistics (v. 22) and R (v. 3.40) were used for the statistical analysis. Kaplan-Meier survival analysis was performed, and Log-rank test was used, with a p-value < 0.05 considered statistically significant. Event-free survival (EFS) was analyzed using death, relapse, resistant disease (blast count >25% at the end of induction), and secondary malignancy as events. Cox proportional hazards models were fitted to estimate the effects of potential risk factors on survival. Chi-squared test was performed on SOX11 expression and clinicopathological prognostic variables. Kruskal-Wallis H and Mann-Whitney U tests were used to evaluate the differential expression of SOX11 in distinct leukemia subgroups (Table S6). All performed statistical tests were two tailed, and no corrections for multiple testing were used.

SOX11 is overexpressed in acute lymphoblastic leukemias. Derailed differentiation and abnormal
proliferation of B-cells are thought to underlie the genesis of BCP-ALL. As TFs are key cell differentiation drivers, we sought to identify aberrantly expressed TFs in BCP-ALL. We utilized a large, curated dataset of microarray-based gene expression profiles retrieved from GEO 34 . This dataset comprises a total of 9,544 hematological gene expression profiles, including 4,418 leukemias, 428 healthy controls, and 862 cell lines. Among the genes with altered expression, TF SOX11 showed prominent expression in MCL and in the E/R and T/P subtypes of BCP-ALL (Fig. 1a). Compared with healthy hematopoietic cells, SOX11 had a 4.7-and 4-fold higher expression in the E/R and T/P subtypes, respectively, with an expression comparable to that in MCL (Fig. 1a). A similar subtype-specific expression of SOX11 was replicated in two additional ALL gene expression datasets (Fig. 1a). Interestingly, the PanALL study revealed that SOX11 expression is also elevated in novel subtypes, such as E/Rlike, IKZF1 N159Y (IKZF1 missense alteration encoding p.Asn159Tyr), MEF2D rearrangement, and DUX4 rearrangement.
As SOX11 belongs to the SoxC family of TFs, we investigated the expression of the two other members of this family. SOX4, which has been reported to affect survival, progression, and proliferation in BCP-ALL 48 , was highly expressed in BCP-ALL and T-ALL, but it showed no subtype-specific expression pattern. The expression of SOX12 did not vary markedly between the studied subtypes (Fig. S1a).
A subtype-specific expression pattern was also present in BCP-ALL cell lines, as REH cells, representing the E/R subtype, and RCH-ACV and 697 cells, representing the T/P subtype, showed an increased expression of SOX11 by RT-qPCR (Fig. 1b). Neither knockdown of E/R in REH cells nor their overexpression in NALM-6 cells had any effect on SOX11 expression levels (Fig. S2a).

Immunohistochemical analysis of SOX11 expression in ALL.
In order to confirm the SOX11 expression at the protein level and relate it to clinical features, we collected a retrospective cohort of 119 B-ALL cases with available bone marrow biopsies and associated clinical data (Table 1). We performed immunohistochemical staining of the biopsy samples by using a SOX11-specific antibody. Staining intensity was graded from 0 to 2, with 0 marking negative, 1 positive, and 2 strongly positive samples. Eighty out of 119 primary B-ALL bone marrow samples stained negative for SOX11, whereas 34 cases were positive and 5 were strongly positive ( Fig. 2a and Table 2). A total of 29/39 (74.4%) of the positively staining B-ALL cases represented either the E/R or T/P subtype. A statistically significant association was observed with the E/R subtype (p-value < 0.001) but not with the T/P subtype (likely because of the low number of cases: N = 4, 3 positive cases). Cases with either KMT2A rearrangement or the Philadelphia chromosome (BCR-ABL1-translocation) did not express the SOX11 protein, and the majority of hyperdiploid cases were also SOX11 negative. Additionally, seven T-ALL cases included in the whole ALL patient cohort were all negative. SOX11 expression is associated with a favorable outcome. We next sought to evaluate the clinical significance of SOX11 expression in B-ALL (N = 119). The overall survival (OS) was better in the SOX11-positive group, and no deaths occurred among SOX11-positive cases (p = 0.039) (Fig. 2b). The EFS and relapse-free survival (RFS) adjusted for the competing event (death) showed similar trends but did not reach statistical significance (Fig. 2b). SOX11 positivity was not associated with good early therapy response, as measured by a minimal residual disease below 0.1% at the end of induction therapy (OR = 0.54, 95% CI 0.22, 1.28, p = 0.17). In multivariate analysis of EFS with covariates (age, WBC, MRD and subtypes), a positive immunohistochemical staining for SOX11 protein showed a favorable trend (Table 3).
When SOX11 immunostaining positivity was analyzed separately within the E/R subtype, SOX11-positive cases had a better OS (Log-rank test p = 0.004), but EFS did not show a statistically significant difference with a hazard ratio of 0.67 (95% CI 0.07, 6.43).
We replicated the survival findings in another dataset. Transcriptomic data from a patient cohort comprising 115 BCP-ALL cases were analyzed for SOX11 expression and patients' survival status 46 . Figure 3a shows that a high SOX11 mRNA expression was associated with a favorable trend in EFS analysis. SOX11 overexpression is associated with DNA hypomethylation. We next investigated the biology behind the increased SOX11 expression in leukemia. Neither primary transcription (GRO-seq, N = 8) nor whole-genome sequencing (WGS, N = 8) of the SOX11 gene in BCP-ALL cases revealed any aberrant enhancer activity or somatic mutations, respectively, in the SOX11 gene or nearby regions (data not shown), prompting us to look for other mechanisms. In MCL, hypomethylation (partly) drives the increased SOX11 expression 30 . We utilized the above-mentioned BCP-ALL patient cohort with readily available genome-wide CpG methylation data 35,46 . Altogether, 23 CpG sites were located within the SOX11 gene locus, and a strong pattern of DNA hypomethylation was seen in patients with a high SOX11 mRNA expression (Fig. 3b,c). Nevertheless, www.nature.com/scientificreports www.nature.com/scientificreports/ DNA hypomethylation of the CpG sites at the SOX11 locus was not associated with a better EFS (Fig. 3a, right  panel). We also tested whether a methyltransferase inhibitor, decitabine, could reverse SOX11 expression in leukemia cell lines. After 72 h of decitabine treatment, a marked increase in SOX11 expression was observed in a concentration-dependent manner in KOPN-8 and REH cells (Figs. 3d and S2b).     www.nature.com/scientificreports www.nature.com/scientificreports/ Knockdown of SOX11 alters gene sets related to cell development, motility, and drug response pathways. We next silenced SOX11 expression in three cell lines that overexpress SOX11 (REH, RCH-ACV, and 697) by using siRNA oligos. Figure 4a,b show that SOX11 expression was decreased to 20-40% at both the mRNA and protein levels compared with scrambled siRNA-transfected cells. Cell viability and proliferation assays did not demonstrate any significant changes (Figs. 4b and S4), and SOX11 knockdown did not have any impact on sensitivity to known leukemia drugs, such as dexamethasone, prednisolone, vincristine, and asparaginase (Fig. S5).
As SOX11 is a TF and it regulates gene expression, we measured changes in gene expression after SOX11 knockdown by using RNA-seq (GSE123943). Three biological replicates were used for each cell line (REH, RCH-ACV, and 697), and the data were analyzed using the R package EdgeR. In GO annotations, many of the altered terms were related to cell migration, adhesion, and differentiation ( Fig. S6 and Supplementary Dataset). On the other hand, GSEA implied the altered expression of MYC and EF2 target genes (Figs. 4c and S6). SOX11 knockdown did not have a significant effect on the expression of other SoxC family members, such as SOX4 and SOX12 (Fig. S7). The Venn diagram in Fig. 4d shows that 18 genes were differentially expressed in all of the three cell lines after SOX11 knockdown (with an adjusted p-value < 0.05). Of these genes, 15 were concordantly down-regulated in all three cell lines and are shown in the heatmap (Fig. 4e, Supplementary Dataset). SOX11 knockdown led to the downregulation of the WD Repeat Domain 1 (WDR1) gene, which is involved in the remodeling of the actin cytoskeleton, regulation of cell migration, motility of neutrophils, and maturation of megakaryocytes 49,50 . Another gene down-regulated by SOX11 knockdown is Midkine (MDK), a secreted growth factor that promotes cell migration and growth and is associated with an adverse prognosis in ALL possibly via increased drug resistance 51,52 . WW Domain Binding Protein 1 Like (WBP1L), also known as OPAL1 (outcome predictor for acute leukemia 1), which was recently identified as a direct target of ETV6 in ALL 53 , was down-regulated by SOX11 silencing. Previous reports on WBP1L's association with a favorable prognosis have since been refuted 54 . However, WBP1L expression had a 2.8-fold increase in the E/R subtype 55 , which coincides with the overexpression of SOX11 in the same subtype and may suggest co-regulation. MLLT11 is involved in lymphoid regulation and is a known partner gene in rare leukemia translocations 56,57 . Similarly, Coiled-coil domain containing 28 A (CCDC28A) is a fusion partner to NUP98 in AML 58 . CANT1 is a calcium-dependent nucleotidase involved in pyrimidine metabolism whose regulation by SOX11 could be related with drug metabolism (cytarabine) and therapy response in ALL 59 . The lipopolysaccharide-induced TNF factor (LITAF) has been suggested to sensitize leukemia cells to chemotherapeutic drugs, especially in cells with a lower expression of LITAF 60 . Both LITAF and Sorting nexin 1 (SNX1) are involved in endosomal trafficking and regulation of cell-surface receptor signaling 61,62 . Taken together, SOX11 knockdown leads to alterations in genes and cellular processes related to leukemia cell motility, adhesion, differentiation, and drug response. SOX11 protein can be detected by flow cytometry. Finally, we searched for flow cytometry markers that could serve as surrogates of SOX11 positivity. In MCL, the surface expression of CD5 is correlated with SOX11 positivity 22 , but in our immunohistochemical staining and microarray dataset 34 , no positive correlation was observed in leukemias (Fig. 5a). Alternative surrogate markers were searched for among the routinely studied cell surface proteins, but none were associated with SOX11 positivity (data not shown). Therefore, we explored the suitability of SOX11 antibodies to discriminate SOX11-positive cases by flow cytometry. Figure 5b shows that the intensity of SOX11 antibody staining could readily separate the high expressors (REH and RCH-ACV) from the low expressors (NALM-6 and KOPN-8) in leukemia cell lines, suggesting that SOX11 antibodies could possibly be used as a biomarker in the future.

Discussion
BCP-ALL is caused by a relatively small number of genetic mutations that impede normal B-cell differentiation, entail self-renewal capacity, and increase proliferative activity. This is evident in transcriptional programs that govern leukemic B-cells. We report here that SOX11, a developmental and neuronal TF 6 , is overexpressed in the E/R and T/P subtypes of BCP-ALL and also in novel E/R-like, IKZF1 N159Y, MEF2D rearrangement, and DUX4 rearrangement subtypes. A high expression is associated with DNA hypomethylation and a favorable clinical outcome. The results suggest that a SOX11-associated transcriptional program is related with a less-aggressive disease and indicates that SOX11 warrants further study as a biomarker for low-risk ALL patients.
We observed the high expression of SOX11 in the E/R and T/P subtypes of BCP-ALL in three separate gene expression datasets at the mRNA level [34][35][36] and confirmed this association at the protein level by immunohistochemistry, confirming previously published data 8,13,29,30 and adding further evidence at the protein level. We showed that the expression may be regulated epigenetically, that is, by hypomethylation of DNA at the SOX11 locus, similar to what was earlier reported for MCL and five cases of BCP-ALL by Vegliante et al. 30 . We did not have material available for epigenetics studies, such as histone modifications, as these could also have a role in SOX11 regulation 30 . Direct manipulation of the E/R fusion did not have any impact on SOX11 expression, suggesting that regulation is indirect.
An interesting finding relates to the clinical significance of SOX11 expression in BCP-ALL, as we observed a better OS in the SOX11-positive cases. A similar trend was also seen in EFS, RFS, and early therapy response. This finding was supported (similar trend) in another dataset with transcriptome expression profiles 46 . In the subgroup analysis of E/R cases, SOX11 positivity retained its prognostic significance, suggesting that (a high) SOX11 expression could possibly be utilized as a biomarker for cases with a very good prognosis. It is noteworthy that our immunohistochemical staining series spans almost two decades, and patients have been treated using several distinct NOPHO ALL chemotherapy protocols 63,64 . As the most recent protocols have conferred the best survival results [63, trying to replicate these findings in the most recent protocols is necessary in the future.
To aid in the screening of SOX11 positivity at diagnosis, we also successfully tested a flow cytometry-based assay in cell lines. SOX11 inclusion into the flow cytometry panel would be convenient compared with immunohistochemical staining of bone marrow biopsies, which is slow and not routinely done in all treatment centers.
SOX11 knockdown did not markedly influence cell viability or proliferation, nor did it affect chemotherapy sensitivity. In MCL, conflicting reports have been made about the effect of knockdown or SOX11 overexpression on cell proliferation and tumor growth [16][17][18][19][20][21][22][23] . As cell viability measurement is a relatively insensitive assay, we performed transcriptional profiling of SOX11 knockdown cells by using RNA-seq and noticed changes in the genes associated with cell migration, adhesion, oxidative phosphorylation, hypoxia, glycolysis, and differentiation, which could explain the association of SOX11 with favorable clinical outcomes. Notably, the changes observed were mostly mild to moderate. There were only few overlapping genes with previous profiling studies in both pro-B-cells 18 and MCL cells 16,17,20,21,65 . For example, we did not see marked changes in the expression of either PAX5, as seen in MCL 21 , or Id1 and Tal1 in pro-B-cells 18 . Interestingly, MDK, which is involved in cell migration and growth, was downregulated by SOX11 knockdown in leukemia cell lines here and in an MCL cell line Z138 16 .
In conclusion, the association of SOX11 expression with a favorable prognosis invites further studies to confirm its prognostic value and applicability as a part of the diagnostic workup.

Data availability
The datasets generated and analyzed in the current study are available in the GEO repository, GSE47051 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE47051), GSE123943 (https://www.ncbi.nlm.nih. gov/geo/query/acc.cgi?acc=GSE123943), and GSE49031 (https://www.ncbi.nlm.nih.gov/geo/query/acc. cgi?acc=GSE49031), or are included in the article or supplementary files. RNA-seq data of the patient samples 46 are not publicly available, as the patient/parent consent does not cover depositing data into repositories; however, they are available from the authors upon reasonable request.