Nuclear factor of activated T-cells, NFATC1, governs FLT3ITD-driven hematopoietic stem cell transformation and a poor prognosis in AML
Acute myeloid leukemia (AML) patients with a high allelic burden of an internal tandem duplication (ITD)-mutated FMS-like Tyrosine Kinase-3 (FLT3) have a dismal outcome. FLT3ITD triggers the proliferation of the quiescent hematopoietic stem cell (HSC) pool but fails to directly transform HSCs. While the inflammatory transcription factor nuclear factor of activated T-cells 2 (NFAT2, NFATC1) is overexpressed in AML, it is unknown whether it plays a role in FLT3ITD-induced HSC transformation.
We generated a triple transgenic mouse model, in which tamoxifen-inducible Cre-recombinase targets expression of a constitutively nuclear transcription factor NFATC1 to FLT3ITD positive HSC. Emerging genotypes were phenotypically, biochemically, and also transcriptionally characterized using RNA sequencing. We also retrospectively analyzed the overall survival of AML patients with different NFATC1 expression status.
We find that NFATC1 governs FLT3ITD-driven precursor cell expansion and transformation, causing a fully penetrant lethal AML. FLT3ITD/NFATC1-AML is re-transplantable in secondary recipients and shows primary resistance to the FLT3ITD-kinase inhibitor quizartinib. Mechanistically, NFATC1 rewires FLT3ITD-dependent signaling output in HSC, involving augmented K-RAS signaling and a selective de novo recruitment of key HSC-transforming signaling pathways such as the Hedgehog- and WNT/B-Catenin signaling pathways. In human AML, NFATC1 overexpression is associated with poor overall survival.
NFATC1 expression causes FLT3ITD-induced transcriptome changes, which are associated with HSC transformation, quizartinib resistance, and a poor prognosis in AML.
KeywordsFLT3ITD Acute myeloid leukemia Hematopoietic stem cells NFATC1 Drug resistance
Acute myeloid leukemia
Age-related clonal hematopoiesis
Bromodomain and extra-terminal protein
Colony forming units
Clonal hematopoiesis of indetermined potential
Common myeloid progenitor
Double transgenic Scl-HSC-Cre-ERT:NFATC1 mouse
- DE genes
Differentially expressed genes
Estrogen receptor, induced by tamoxifen
Transgenic Flt3ITD mouse
Triple transgenic Flt3ITD:Scl-HSC-Cre-ERT:NFATC1 mouse
Gene set enrichment analysis
Hematopoietic stem cells
Mouse Linnegc-kitpos progenitor cells
Mouse LinnegSca1posc-kitpos stem cells
The cancer genome atlas project
Transcription factor targets
Tyrosine kinase inhibitor
Wild type mouse
Acute myeloid leukemia (AML) is mainly a fatal disease. AML evolution usually begins with sentinel de novo acquisition of mutations in epigenetic modifier genes such as DNMT3A, ASXL1, or TET2, which also characterize clonal hematopoiesis of healthy individuals, referred to as CHIP or age-related clonal hematopoiesis, ARCH . Progression from CHIP to AML is frequently associated by the consecutive acquisition of additional mutations such as in NPM1, RAS, KIT, or FLT3 . However, recent evidence also shows that clonal evolution from CHIP to AML is not exactly a linear path . In fact, the requirements, under which CHIP mutations become myeloid driver mutations, cooperate with other genetic or epigenetic changes to eventually cause AML, are not well understood. De novo chromatin alterations that are caused by aberrant transcription factor (TF) signaling could be very critical in this process . One such TF is the nuclear factor of activated T-cells (NFAT). NFATs are a family of TFs that are physiologically activated by the calcium-activated serine/threonine phosphatase calcineurin, which dephosphorylates NFAT, thereby leading to their nuclear translocation and target gene regulation . NFAT proteins play a pivotal role during T-cell maturation and activation . Aberrant NFAT signaling is causally involved in the development of chronic lymphocytic leukemia, non-Hodgkin lymphoma, pancreatic cancer, and several other malignancies. We have previously shown that nuclear NFATC1-induced epigenetic changes convert benign Ras-induced pancreatic adenomas into rapidly lethal, metastatic cancers . NFAT proteins are expressed in normal and malignant lymphatic hematopoiesis. We showed recently that NFATC1 is overexpressed in AML where it mediates resistance to sorafenib in vitro . Sorafenib is a potent inhibitor of the mutated FLT3-receptor tyrosine kinase variant, FLT3ITD, which is frequently found in normal karyotype AML [2, 9]. FLT3ITD constitutively activates a signaling network, including the RAS- and STAT-signaling pathways. FLT3ITD confers a poor risk in AML [10, 11]. This provided the rationale for the development of FLT3-specific TKI in order to improve outcome. However, although FLT3-TKI therapy proved to be clinically effective, responses are usually temporary through the development of TKI resistance [12, 13, 14, 15, 16]. Notably, although FLT3ITD is a critical driver of AML pathogenesis and dictates prognosis, it fails to transform normal stem cells into self-renewing AML in vivo [17, 18, 19] unless other oncogenes are co-expressed [20, 21]. This is because FLT3 ITD-signaling causes overproliferation of normally quiescent stem cells, leading to stem cell depletion rather than leukemic transformation .
Here, we show for the first time that NFATC1, a key regulator of inflammatory responses , potently synergizes with FLT3ITD, by causing stem cell expansion rather stem cell depletion, resulting in the development of a FLT3-kinase inhibitor-resistant lethal AML. In human AML, overexpression of NFATC1 can be linked to a poor prognosis, supporting the critical importance of NFATC1 in AML biology.
Nuclear NFATC1 induces a rapidly lethal FLT3ITD-driven leukemia
NFATC1 increases FLT3ITD-dependent clonogenic cell mass through precursor and stem cell expansion
FCN leukemias were transplantable. When 106 and 107 FCN leukemia cells were injected into sub-lethally irradiated WT mice, which were sacrificed after 3 months of observation. FCN engraftment was seen in 1/3 and 3/3 animals, respectively (Fig. 2f), while no engraftment was seen in mice that received 103 and 105 cells. Recipient animals developed a monocytic leukemia and splenomegaly (Fig. 2e). Importantly, in contrast to secondary F leukemias, which showed less LSK and LK than primary F leukemias (Fig. 2h, i), as exemplary shown (Additional file 3: Figure S3), secondary recipients of FCN leukemias fully re-established the FCN leukemia phenotype, maintained or even increased the proportions of LSK, LK, and monocytic fractions (Fig. 2h–k).
NFATC1 causes primary resistance to the FLT3-inhibitor quizartinib
De novo recruitment of NFATC1 dependent leukemic signaling pathways in the context of FLT3ITD
NFATC1 and FLT3-signaling govern poor survival in AML
Here, we provide evidence that NFATC1 is an essential driver of FLT3ITD-dependent leukemic transformation and demonstrate a previously unknown prognostic role for NFATC1 in AML. By generating a transgenic Flt3ITD-mouse model with tamoxifen-inducible targeted expression of NFATC1 in HSC, we show that expression of nuclear NFATC1 is a key factor driving FLT3ITD-induced stem cell expansion and transformation and confers primary FLT3ITD inhibitor resistance.
Development of this mouse model was prompted by our recent observations showing that many FLT3ITD-positive and -negative AML patients frequently overexpress NFATC1. We have also shown previously in a murine model of inflammation-driven pancreatic cancerogenesis that NFATC1 is a relay switch converting KrasG12D-driven pre-neoplastic changes into a rapidly lethal metastatic pancreatic cancer . Here, we demonstrate that expression of the nuclear NFATC1 in stem cells changes the biological outcome of aberrant FLT3ITD kinase signaling. In the presence of NFATC1, FLT3ITD-signaling leads to extensive precursor cell expansion, development of a fully penetrant lethal and quizartinib-resistant AML. In contrast, neither expression of FLT3ITD nor NFATC1 alone in HSC causes an apparent phenotype in vivo.
Fundamental gene expression changes underlie the biological differences among F, CN, and FCN animals. A GSEA-TFT-based alignment of differential gene expression patterns and target motifs for TFs in the promoters/enhancers of DE genes strongly suggested that there is qualitative and quantitative cooperativity between NFATC1 and FLT3ITD. Qualitative cooperativity is presumably due to an NFATC1-dependent de novo recruitment of TFs to promoters in FCN, but not F or CN DE genes. Among them were TFs that are known to be involved in stem cell transformation such as LEF1 , HOXA4  or FLT3ITD-dependent AML transformation, such as RUNX1 , STAT5 [28, 34], and MYB . In line with these qualitative changes in gene expression, key self-renewal pathways of leukemogenesis, such as the Hedgehog- and WNT/B-Catenin signaling pathways, are recruited de novo only in FCN stem cells. Both these pathways were shown to be essential for the maintenance of cancer stem cells in myeloid leukemias and can be specifically targeted by therapeutic intervention [36, 37].
According to our GSEA TFT analysis, NFATC1 enables the recruitment of FLT3ITD-inducible TFs such as the MAPK-inducible TFs AP1 and NF-κB, STAT5 , and FOXO4  to significantly more target genes than would be accessible with FLT3ITD expression alone. RNA sequencing indeed suggests a dramatic amplification of FLT3ITD-signaling output in the presence of nuclear NFATC1, that is, recruitment of more genes into existing FLT3ITD-dependent gene sets. Intriguingly, GSEA-TFT analysis of FCN stem cells shows activation of known NFATC1 binding partners AP1 and RUNX1 [39, 40], which were recently uncovered as key drivers of an open chromatin structure in FLT3ITD-positive primary AML blasts . This supports that NFATC1’s transforming potential in the context of oncogenic FLT3ITD-signaling is linked to its capability to induce or promote active chromatin states [7, 39, 42].
Chromatin-modifying roles of TFs are not unprecedented. They have been reported for TF involved in leukemogenesis such as TCF/LEF1  or FOX . Our data suggest that NFATC1/FLT3ITD cooperativity leads to a gain of transcriptional plasticity, which is an important prerequisite for the acquisition of increased fitness and thus development of phenotypic diversity such as drug resistance under selective pressure  (Additional file 5: Figure S5).
In contrast to F animals, FCN animals were resistant to a clinically effective FLT3ITD-inhibitor, quizartinib , because only F animals showed a significant reduction of LK and LSK. The fact that quizartinib did not more potently eliminate F leukemia might be explained by the fact that in contrast to human FLT3ITD positive AML, FLT3ITD fails to induce oncogenic dependence in FLT3ITD-induced myeloproliferation in F mice. In AML patients, FLT3ITD induces oncogene dependence as the basis of therapy response to FLT3 inhibition , yet remissions are short lived and mostly not curative [13, 15, 46]. We have previously shown that NFATC1-induced FLT3ITD-inhibitor resistance is FLT3ITD-independent . In contrast to FLT3ITD, bromodomain and extra-terminal (BET) proteins, such as BRD4, which read chromatin and regulate transcription, are non-oncogene addiction targets in MLL-AF9-positive AML . BET inhibitor resistance is mediated via activation of WNT signaling, which involves a profound transcriptional reprogramming . Thus, it is tempting to speculate that WNT-signaling activation through FLT3ITD/NFATC1 cooperativity (Fig. 5d) mediates FLT3ITD-independent FLT3-inhibitor resistance that could be overcome using BET- or WNT-signaling inhibitors.
Moreover, the recruitment of independent oncogenic signaling pathways through FLT3ITD/NFATC1 cooperativity might explain, why oncogenic driver mutations such as FLT3ITD—though being instrumental for AML initiation—become dispensable for AML maintenance, once NFATC1 is overexpressed. The importance of NFATC1 overexpression for AML transformation as suggested by the FCN mouse model is supported by the observation of a dismal prognostic influence of NFATC1 overexpression in AML. Ongoing studies will reveal whether pre-emptive blocking of inflammatory signals that trigger NFATC1 expression or NFATC1-activation itself using cyclosporine could be a concept to prevent NFATC1-dependent leukemogenesis.
Flt3ITD mice (B6.129 background) were bought from Jackson Laboratory, transgenic NFATC1 mice and conditional Cre-SCL-HSC-ER mice (both C57Bl/6j background) generated as described previously [7, 24] (for breeding scheme, see Additional file 1: Figure S1). All mice (except recipient mice in transplantation experiment) were treated with CreActive TAM400 (LASvendi LASCRdiet) tamoxifen diet for 21 days. All animal procedures were conducted in accordance with TierSchG and were approved by RP Giessen.
A total of 50 μl of blood was taken from the tail vein and max of 10% (v/v) of Di-Sodium-EDTA 1.107% solution was added to prevent coagulation. Blood films were stained with Wright-Giemsa method. For blood cell counts, a dilution of 1:7 in NaCl 0.9% was made and measured in a capillary mode on SYSMEX XS 1000i.
See Additional file 6: Table S1 for primer and PCR reaction information.
Cytospins and pathology
Cytospins were prepared with 2 × 105 cells and stained with Wright-Giemsa. For pathology, mouse organs were fixed in 4% PFA (paraformaldehyde) and stained with H&E.
Blood samples for Western blotting were obtained from the tail vein and erythrocytes were lysed. Total protein lysates and transfer were performed as described previously . The following primary antibodies were used: mouse monoclonal anti-HA, clone 6E2 (Cell Signaling, Danvers, USA), mouse monoclonal anti-beta-actin, and clone AC-15 (Sigma-Aldrich, Steinheim, Germany). Secondary goat anti-mouse horseradish peroxidase (HSP)-conjugated antibody was used (Dako Cytomation, Glostrup, Denmark) (see Additional file 7: Table S2).
Sections of paraffin-embedded bone marrow were immunohistochemically stained using the NovolinkTM Polymer Detection Kit (Leica, Wetzlar, Germany), pH 6.0 for antigen unmasking and the NFAT2 monoclonal antibody, clone ab25916 (Abcam, Cambridge, MA, USA) (see Additional file 7: Table S2).
Flow cytometry and cell sorting
Flow cytometry analysis was performed as previously described . Cell sorting was performed using the same antibodies and staining protocols (for the list of antibodies used, see Additional file 7: Table S2).
In vitro colony forming assays
Total bone marrow after erythrocyte lysis or sorted populations were seeded into full cytokine methylcellulose medium (Methocult M3434, STEMCELL Technologies) at 4000 cells for total bone marrow or 500 cells for sorted LSK or LK per well, respectively. Colonies were counted at 7 to 14 days. Cells were re-plated at the same cell numbers.
Bone marrow transplantation
Total bone marrow from diseased animals or WT controls was isolated, erythrocytes were lysed and four different concentrations of total cells (1 × 107, 1 × 106, 1 × 105, and 1 × 103) transplanted via tail vein injection into sub-lethally (7 Gy) irradiated WT recipients. According to the mouse experimental proposal, transplanted recipient mice were sacrificed and analyzed 3 months after transplantation.
In vitro and in vivo treatment
For in vitro cytarabine treatment, total bone marrow cells were seeded after erythrocyte lysis into 6-well plates at 4 × 104 cells per well in Iscove’s Modified Dulbecco’s Medium (plus 20% fetal calf serum, 2% penicillin/streptomycin and cytokines: IL-3, IL-6, stem cell factor 0.1%) and treated with 0.9% sodium chloride solution or cytarabine at concentrations of 10 nM or 100 nM for 24 h. Cells were washed with phosphate-buffered saline and seeded into methylcellulose medium in triplicates at 4 × 103 cells per dish (as described [8, 49]). The colony number was evaluated after 7–10 days.
For in vivo treatment, the AC220 treatment group received AC220 daily at 10 mg/kg, suspended in 5% 2-hydroxypropyl-ß-cyclodextrin for 21 days. The chemotherapy group received intraperitoneal injections of 100 mg/kg cytarabine for 5 days and 3 mg/kg doxorubicin for 3 days. Both control groups received solvent.
RNA sequencing and analysis
Total RNAs were isolated from sorted mouse bone marrow stem cells (LSK, n = 2 to 3 per genotype) using RNeasy Micro Kit (Qiagen), according to the manufacturer’s protocol. For WT and CN genotypes, several bone marrow samples were pooled together to get enough material per one mRNA sample. RNA libraries were generated with the NEBNext ® UltraTM RNA Library Prep Kit for Illumina (New England Biolabs® Inc.), according to the protocol of the manufacturer. The libraries were sequenced on an Illumina HiSeq 2000, requested Read Length HiSeq 2 × 75 bp with paired end.
Mapping was done with bwa aligner to GRCm38 (mm10) mouse genome (see Additional file 8: Table S3 for raw counts). DESeq2 was used to calculate differentially expressed genes. DE genes were filtered by FDR adjusted p value less than 0.01 and log2 fold change greater than 1 and less than − 1 (see Additional file 9: Table S4 for DE genes). The information on the subcellular localization of the proteins coded by the DE genes was collected from the databases http://www.uniprot.org and https://imb.uq.edu.au. GSEA analysis  was performed with the filtered DE genes on the following platform: http://software.broadinstitute.org/gsea/msigdb/annotate.jsp. We used H, C3 TFT, or C2 MSigDB databases and considered only gene sets with FDR q value less than 0.01.
AML patient data analysis
Overall survival of AML patients was analyzed according to NFATC1 expression levels. Kaplan Meier plots were generated for the data sets GSE21417 and TCGA using SurvExpress—a biomarker validation tool and database for cancer gene expression data . The median NFATC1 expression level was used as the threshold to split AML patients into two high versus low NFATC1-expressing cohorts.
Data, which followed normal distribution, are presented as mean ± SD. Differences between groups were analyzed by one-way ANOVA or two-way ANOVA with Bonferroni or Dunnett’s multiple comparison post testing; p < 0.05 was considered significant. The data, which did not follow normal distribution, are presented as median ± IR. The differences between groups were analyzed by Mann-Whitney non-parametric test. Log-rank test was used to calculate the p value of the difference in the OS.
For the first time, it is shown that NFATC1—a transcription factor that is induced upon inflammatory stimuli—converts FLT3ITD-induced myeloproliferation into an FLT3ITD inhibitor-resistant, lethal AML, which is associated with NFATC1-dependent de novo recruitment of key oncogenic signaling pathways. This implies that therapeutic targeting of these pathways, NFATc1 itself using calcineurin inhibitors or inflammation could be a strategy to treat NFATc1-expressing myeloid neoplasias.
We would like to acknowledge the assistance of the Irradiation Core Facility of the Philipps University supported in part by the DFG, in particular Prof. Dr. K. Zink, Prof. Dr. R.Engenhart-Callibic, and Dr. A. Arenz. We thank PD Dr. C. Brendel, G. Giel, and Dr. H. Reifer for help with cell sorting. We also thank Prof. Dr. R. Moll and V. Wischmann for assistance with hisology. This work was supported by the Deutsche José Carreras Leukämiestiftung (AR12/12). It was also supported by the Deutsche Forschungsgemeinschaft, DFG, Klinische Forschergruppe 210 “Genetics of Drug resistance in Cancer” (TP7 to AB), and an unrestricted grant of the “Anneliese Pohl Stiftung” (to AB). We thank Sonja Tajstra for her excellent technical assistance in performing mice work.
MS helped in project design, performed the experiments, performed bioinformatics analysis, analyzed data, and helped in writing the manuscript. YW designed and performed RNA sequencing experiments, analyzed the data, and reviewed the manuscript. CM helped in performing experiments, helped in the design of the project, analyzed the data, and reviewed the manuscript. KHM and TH designed experiments, analyzed data, and contributed in writing the manuscript, JRG, VE, and EH helped in the design of mice experiments, analyzed the data, and reviewed the manuscript. SG designed experiments, performed immunohistochemistry, analyzed the data, and reviewed the manuscript. DP and VB designed RNA-sequencing experiments, helped in performing sequencing, helped in bioinformatic analysis, and reviewed the manuscript. OR performed bioinformatic analysis of sequencing data and reviewed the manuscript. AB designed and coordinated the project, analyzed the data, and wrote the paper. The authors declare that they have no competing interests. All authors read and approved the final manuscipt.
This work was funded by DFG (KFO210/TP7) and DCLS (AR12/12).
All animal procedures were conducted in accordance with TierSchG and were approved by RP Giessen (approval number V54-19c2015h01MR20/36Nr.17/2013).
Consent for publication
The authors declare that they have no competing interests.
- 3.Chen J, Kao Y-R, Sun D, Todorova TI, Reynolds D, Narayanagari S-R, et al. Myelodysplastic syndrome progression to acute myeloid leukemia at the stem cell level. Nat Med. Nature Publishing Group. 2018:1.Google Scholar
- 7.Baumgart S, Chen N-M, Siveke JT, König A, Zhang J-S, Singh SK, et al. Inflammation-induced NFATc1-STAT3 transcription complex promotes pancreatic cancer initiation by KrasG12D. Cancer Discov. Am Assoc Cancer Res. 2014;4:688–701.Google Scholar
- 13.Metzelder SK, Schroeder T, Finck A, Scholl S, Fey M, Götze K, et al. High activity of sorafenib in FLT3-ITD-positive acute myeloid leukemia synergizes with allo-immune effects to induce sustained responses. Leukemia. 2012.Google Scholar
- 17.Li L, Piloto O, Nguyen HB, Greenberg K, Takamiya K, Racke F, et al. Knock-in of an internal tandem duplication mutation into murine FLT3 confers myeloproliferative disease in a mouse model. Blood. Am Soc Hematol. 2008;111:3849–58.Google Scholar
- 20.Behrens K, Maul K, Tekin N, Kriebitzsch N, Indenbirken D, Prassolov V, et al. RUNX1 cooperates with FLT3-ITD to induce leukemia. J Exp Med. Rockefeller University Press. 2017;214:737–52.Google Scholar
- 21.Greenblatt S, Li L, Slape C, Nguyen B, Novak R, Duffield A, et al. Knock-in of a FLT3/ITD mutation cooperates with a NUP98-HOXD13 fusion to generate acute myeloid leukemia in a mouse model. Blood. Am Soc of Hematol. 2012;119:2883–94.Google Scholar
- 27.Petropoulos K, Arseni N, Schessl C, Stadler CR, Rawat VPS, Deshpande AJ, et al. A novel role for Lef-1, a central transcription mediator of Wnt signaling, in leukemogenesis. J Exp Med. Rockefeller University Press. 2008;205:515–22.Google Scholar
- 28.Choudhary C, Brandts C, Schwable J, Tickenbrock L, Sargin B, Ueker A, et al. Activation mechanisms of STAT5 by oncogenic Flt3-ITD. Blood, vol. 110: Am Soc Hematol; 2007. p. 370–4.Google Scholar
- 29.Metzeler KH, Hummel M, Bloomfield CD, Spiekermann K, Braess J, Sauerland MC, et al. An 86-probe-set gene-expression signature predicts survival in cytogenetically normal acute myeloid leukemia. Blood, vol. 112: Am Soc Hematol; 2008. p. 4193–201.Google Scholar
- 31.Aguirre-Gamboa R, Gomez-Rueda H, Martínez-Ledesma E, Martínez-Torteya A, Chacolla-Huaringa R, Rodriguez-Barrientos A, et al. SurvExpress: an online biomarker validation tool and database for cancer gene expression data using survival analysis. Cho WCS, editor, vol. 8: PLoS ONE. Public Library of Science; 2013. p. e74250.Google Scholar
- 35.Hess JL, Bittner CB, Zeisig DT, Bach C, Fuchs U, Borkhardt A, et al. c-Myb is an essential downstream target for homeobox-mediated transformation of hematopoietic cells. Blood. American Society of Hematology. 2006;108:297–304.Google Scholar
- 40.Macián F, López-Rodríguez C, Rao A. Partners in transcription: NFAT and AP-1. Oncogene. Nature Publishing Group. 2001;20:2476–89.Google Scholar
- 45.Cortes JE, Tallman MS, Schiller GJ, Trone D, Gammon G, Goldberg SL, et al. Phase 2b study of 2 dosing regimens of quizartinib monotherapy in FLT3-ITD-mutated, relapsed or refractory AML. Blood, vol. 132: American Society of Hematology; 2018. p. 598–607.Google Scholar
- 46.Cortes JE, Kantarjian H, Foran JM, Ghirdaladze D, Zodelava M, Borthakur G, et al. Phase I study of quizartinib administered daily to patients with relapsed or refractory acute myeloid leukemia irrespective of FMS-like tyrosine kinase 3-internal tandem duplication status. J Clin Oncol. 2013;31:3681–7.CrossRefGoogle Scholar
- 47.Zuber J, Shi J, Wang E, Rappaport AR, Herrmann H, Sison EA, et al. RNAi screen identifies Brd4 as a therapeutic target in acute myeloid leukaemia. Nature. Nature Publishing Group. 2011;478:524–8.Google Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.