Deletion of HP1γ in cardiac myocytes affects H4K20me3 levels but does not impact cardiac growth
Heterochromatin, which is formed when tri-methyl lysine 9 of histone H3 (H3K9me3) is bound by heterochromatin 1 proteins (HP1s), plays an important role in differentiation and senescence by silencing cell cycle genes. Cardiac myocytes (CMs) accumulate heterochromatin during differentiation and demethylation of H3K9me3 inhibits cell cycle gene silencing and cell cycle exit in CMs; however, it is unclear if this process is mediated by HP1s. In this study, we created a conditional CM-specific HP1 gamma (HP1γ) knockout (KO) mouse model and tested whether HP1γ is required for cell cycle gene silencing and cardiac growth.
HP1γ KO mice were generated by crossing HP1γ floxed mice (fl) with mice expressing Cre recombinase driven by the Nkx2.5 (cardiac progenitor gene) promoter (Cre). We confirmed that deletion of critical exons of HP1γ led to undetectable levels of HP1γ protein in HP1γ KO (Cre;fl/fl) CMs. Analysis of cardiac size and function by echo revealed no significant differences between HP1γ KO and control (WT, Cre, fl/fl) mice. No significant difference in expression of cell cycle genes or cardiac-specific genes was observed. Global transcriptome analysis demonstrated a very moderate effect of HP1γ deletion on global gene expression, with only 51 genes differentially expressed in HP1γ KO CMs. We found that HP1β protein, but not HP1α, was significantly upregulated and that subnuclear localization of HP1β to perinuclear heterochromatin was increased in HP1γ KO CMs. Although HP1γ KO had no effect on H3K9me3 levels, we found a significant reduction in another major heterochromatin mark, tri-methylated lysine 20 of histone H4 (H4K20me3).
These data indicate that HP1γ is dispensable for cell cycle exit and normal cardiac growth but has a significant role in maintaining H4K20me3 and regulating a limited number of genes in CMs.
KeywordsHP1γ Conditional knockout Cardiac myocytes Cell cycle Gene expression H4K20me3 H3K9me3
adult cardiac myocyte
senescence-associated heterochromatin foci
tri-methyl lysine 9 of histone H3
tri-methyl lysine 20 of histone H4
heterochromatin protein 1
histone methyl transferase
chromo shadow domain
embryonic cardiac myocyte
transverse aortic constriction
Adult cardiac myocytes (ACMs) are terminally differentiated cells that express a unique cell-specific transcriptional profile. During terminal differentiation of CMs, cell cycle genes are irreversibly silenced [1, 2] by a mechanism which we have hypothesized is similar to what occurs in cellular senescence [3, 4]. Induction of senescence is thought to involve the sequestration of cell cycle genes within senescence-associated heterochromatin foci (SAHF), which are enriched with heterochromatic markers such as tri-methyl lysine 9 of histone H3 (H3K9me3), tri-methyl lysine 20 of histone H4 (H4K20me3), macroH2A and heterochromatin protein 1 (HP1) [5, 6]. For example, E2F target genes that are required for proliferation localize to SAHF  and are enriched with heterochromatic marks . Because heterochromatin acts as a physical barrier, cell cycle genes are thought to no longer be accessible to transcriptional machinery that drive expression and thereby lead to exit of the cell cycle.
Heterochromatin assembles not only into large domains of constitutive heterochromatin but can also regulate gene activity on a gene-by-gene basis. For example, the master cell cycle regulator retinoblastoma protein (RB) [8, 9] recruits key players of heterochromatin formation such as Suv39h1 H3K9 histone metyltransferase (HMT), HP1s and histone deacetylases [4, 5, 10, 11, 12] to E2F target genes and assemble localized heterochromatin complexes. Indeed, loss of RB impaired SAHF formation . RB is one of member of gene family encoding three structurally and functionally similar proteins, RB, p107 and p130. In ACMs, RB and p130 are expressed [1, 4]. We previously knocked out both RB and p130 in ACMs and found that RB/p130 double KO (RB/p130-DKO) ACMs failed to silence cell cycle genes and reentered cell cycle . In addition, cardiac-specific H3K9me3 depletion prevented both cell cycle exit and terminal differentiation . These results indicate that heterochromatin formation through H3K9me3 could mediate cell cycle exit in CM terminal differentiation.
In this study, we tested the hypothesis that HP1γ is required for the cell cycle gene silencing in CMs. Accordingly, we generated a novel conditional mouse model where HP1γ was specifically deleted in CMs. Surprisingly, loss of HP1γ had minimal effect on cardiac gene expression and differentiation.
Creation of Nkx2.5-Cre;HP1γfl/fl
The HP1γfl mouse was generated from the HP1γhypo/hypo  by excision of the FRT-NeO-FRT cassette after crossing to a FLPe mouse . The resulting HP1γfl mouse was crossed to the Nkx2.5-Cre mouse  to achieve cardiac-specific HP1γ KO. Litter mates were used as controls for this study. Mice were housed in a temperature-controlled environment with 12 h light/dark cycles where they received food and water ad libitum. All protocols concerning the use of animals were approved by the Institutional Animal Care and Use Committees at University of Washington.
Cardiac myocyte isolation
Embryonic CMs were isolated by enzymatic digestion. Briefly, E15-timed pregnant female mice were euthanized and embryonic hearts were extracted from embryos and pooled into ice-cold Ads buffer (116 mM NaCl, 20 mM HEPES, 11 mM NaH2PO4, 5.5 mM Glucose, 5.4 mM KCl, 0.83 mM MgSO4, pH 7.4) supplemented with 5 U/ml heparin. Hearts were washed and atriums were removed in Ads buffer with heparin. The ventricles were minced and digested with 5 mg/ml collagenase Type II (Worthington, 4176) and 1 mg/ml pancreatin (Sigma, P3292) prepared at 37 °C for 20 min. Cardiac myocytes were purified using a Percoll gradient, resulting in ~ 90% pure CMs.
Adult CMs were isolated by using Langendorff perfusion digestion. Briefly, adult hearts were washed by perfusing with Ca2+ free Tyrode’s buffer (126 mM NaCl, 5.4 mM KCl, 0.33 mM NaH2PO4, 1 mM MgCl2, 10 mM HEPES, 10 mM Glu, 20 mM taurine, pH 7.4) supplemented with 25 μM blebbistatin (Toronto Research Chemicals Inc. and BOC Sciences) for 1–2 min and then enzymatically digested at 37C by perfusing with 0.69 U/ml Liberase TH (Roche) prepared with blebbistatin supplemented-Ca2+ free Tyrode’s buffer for 8–12 min. Hearts were mechanically dissociated in ice-cold KB solution (20 mM KCl, 10 mM KH2PO4, 70 mM potassium glutamate, 1 mM MgCl2, 25 mM Glucose, 20 mM Taurine, 0.5 mM EGTA, 10 mM HEPES, 0.1% albumin, pH 7.40). CM suspensions were passed through a 100–200 μm cell strainer to remove tissue debris and then purified by low-speed centrifugation (50×g for 1 min) 3 times, resulting in ~ 90% pure ACMs.
Nuclear extracts were prepared from isolated CMs
Purified ACMs were resuspended in NP40 lysis buffer (0.5% NP40, 25 mM KCl, 3 mM MgCl2, 10 mM Tris–HCl, pH 8.0) and homogenized until nuclei were released from the cytoskeleton. Extraction buffer was supplemented with 1 mM Na3VO, 1 mM NaF, 1 mM phenylmethylsulfonyl fluoride and 1× Protease Inhibitor Cocktail (Millipore 539134). The homogenate was centrifuged with 50×g for 1 min at 4 °C and supernatant (enriched nuclear) was corrected. The enriched nuclear was pelleted with 800×g for 10 min at 4 °C. Nuclear pellets were resuspended in SDS/βME nuclear buffer (1% SDS, 25 mM 2-mercaptoethanol, 137 mM NaCl, 0.5% NP40, 25 mM KCl, 3 mM MgCl2, 10 mM Tris–HCl, pH 8.0) and sonicated with probe sonicator (Qsonica, CL-18) with 25% power for 10 s 2 time (30 s interval). Nuclear extracts were cleared by centrifugation at 20,000×g for 10 min at 15 °C. DNA concentration was measured using Quant-iT™ PicoGreen® dsDNA Reagent (Life technology) and used to normalize samples. Nuclear extract (50–250 ng DNA) were separated by SDS–PAGE, transferred to polyvinylidene fluoride membrane and probed using specific primary antibodies and appropriate HRP-conjugated secondary antibody for ECL detection. Antibodies used are listed in Additional file 3.
Isolated CMs were fixed with cold methanol. Heart tissues were fixed with 4% PFA and paraffin- embedded. Fixed CMs and paraffin-embedded tissue sections were immunostained following standard protocols. Briefly, fixed CMs and deparaffinized/rehydrated tissues were blocked with 1% bovine serum albumin (BSA)/phosphate buffered saline (PBS). Primary antibodies were diluted in 1% BSA/PBS and incubated with blocked CMs and tissues for overnight at 4 °C. Appropriate secondary antibodies conjugated with Alexa Fluor® Dyes (Life Technologies) were diluted in 1% BSA/PBS and incubated with primary antibodies-probed CMs and tissues for 1 h at room temperature. Nuclei were counter stained with Hoechst (Life Technologies). Where indicated, heart tissues were stained with wheat germ agglutinin (WGA) conjugated with Oregon Green™ 488 (Thermo) to visualize cell border. Fluorescence images were acquired using a confocal microscope (Nikon A1R).
Total RNA was extracted from isolated CMs. All total RNA were treated with DNase I following manufacture’s protocol to avoid genomic DNA contamination. Embryonic CMs total RNA were extracted using RNeasy Micro Kit (Qiagen). Adult CMs total RNA was pre-cleared using TRIzol™ Reagent (Invitrogen) and then purified using RNeasy Micro Kit (Qiagen). 1 μg of total RNA was converted to cDNA using Transcriptor First Strand cDNA Synthesis Kit (Roche). cDNA corresponding to 5 ng of total RNA was used for quantitative PCR. Quantitative PCR was performed using SYBR™ Select Master Mix (Life Technologies). Primer sequences used in this study are listed in Additional files 2 and 3. PCR specificity was confirmed by amplicon dissociation curve and by running amplicon on agarose gel. Standard curves were created for each assay. Target gene expression was normalized against internal control S26 or GAPDH and presented as relative expression value against control sample as described in figure legend after confirming expression was similar between the genotypes or operations we used in this study.
Transverse aortic constriction (TAC) surgery
8 week baseline RNA-seq: Total RNA were extracted from 3 HP1γ KO CMs and 3 controls CMs (WT, Cre and fl/fl, one of each) at 8 week old. Library construction and sequencing was performed using commercial service (Omega Bioservices). Briefly, sequenced reads were mapped on the mouse genome (mm10) assembly using TopHat . Transcripts were assembled and fragments per kilobase of transcript per million (FPKM) was calculated by Cufflinks . Differential gene expression was calculated using Cuffdiff . We defined differentially expressed gene as false discovery rate (FDR) < 0.05 and fold change ≥ 2.
TAC-Operated CMs RNA-seq: Three independent samples per group were used: Sham-fl/fl, Sham-Cre;fl/fl, TAC-fl/fl and TAC-Cre;fl/fl. Library construction and sequencing was performed by the Genomics Core Service at Fred Hutchinson Cancer Research Center (Seattle, WA). Sequenced reads were mapped on mouse mm10 using TopHat, counts for each gene were generated using HTSeq , and normalization and comparison of gene expression difference were conducted by edgeR [31, 32]. Genes with FDR < 0.05 and fold change ≥ 2 are defined as significantly differentially expressed.
HP1γ is the major HP1 isoform in ACM
To determine if heterochromatin accumulates in CMs during terminal differentiation (Fig. 1a) CMs from E15 embryo and 10-week adult mouse hearts were stained with Hoechst dye, which stains AT-rich DNA that is concentrated within heterochromatin . E15 embryonic CMs (ECMs) showed a diffuse staining pattern with multiple small heterochromatin foci. In contrast, ACMs showed accumulation of heterochromatin at the nuclear periphery with a few large internal heterochromatin foci.
To explore the role of HP1s in heterochromatin accumulation in CMs further we characterized expression of HP1s during cardiac development. We performed RT-qPCR using purified CMs to compare expression of each HP1 isoform (Fig. 1b). All three HP1 isoforms are expressed in both ECM and ACM. Of these isoforms, HP1γ demonstrated the highest gene expression in both ECM and ACM (1.9-fold vs HP1α; p = 0.0016 and 4.8-fold vs HP1β; p = 0.0002 in ECM, 1.8-fold vs HP1α; p = 0.0165 and 3.1-fold vs HP1β; p = 0.0013 in ACM).
Next, we compared the nuclear localization pattern of HP1s (Fig. 1c). In ECMs, HP1α and HP1β are localized with heterochromatin foci (Hoechst), but HP1γ showed diffuse staining which is euchromatic pattern. In contrast, in ACMs, HP1α and HP1β staining was speckled and no longer confined to heterochromatin foci, while HP1γ staining pattern overlapped with heterochromatin, which is similar to H3K9me3 staining. Co-staining HP1s with H3K9me3 further confirmed that HP1γ is localized with H3K9me3 coincident with heterochromatin foci in ACMs (Additional file 1: Fig. S1). These data demonstrate that HP1γ is the major isoform of HP1s that co-localizes with H3K9me3 in ACMs, suggesting it may play a role in heterochromatin organization during terminal differentiation of CMs.
Creation of cardiac myocyte-specific HP1γ KO mice
Loss of HP1γ causes reduction in H4K20me3 and upregulation of HP1β
Next, we tested if loss of HP1γ effected the other isoforms of HP1s, HP1α and HP1β. HP1α protein level was unchanged between controls and HP1γ KO (Fig. 3a). However, HP1β protein expression was 2.3 times higher in HP1γ KO CM compared to controls (p < 0.01 vs WT, Cre and fl/fl; Additional file 5: Fig. S3). Immunofluorescence staining demonstrated that there is no difference of HP1α localization pattern or signal intensity. In contrast, HP1β was increased and changed in localization to the nuclear periphery (Fig. 3c and Additional file 6: Fig. S4B). Thus, loss of HP1γ affects HP1β expression and localization, but not HP1α.
HP1γ depletion has no effect on cell cycle gene expression and cardiac growth
Impact of HP1γ loss on global gene expression
A limited effect of HP1γ KO on hypertrophic growth of heart
HP1γ loss does not alter gene expression profile after hemodynamic stress
Our previous data suggested that H3K9me3 controlled cell cycle gene silencing in CMs. In this study, we tested whether H3K9me3 reader proteins, HP1s, could mediate this process. We focused on HP1γ for several reasons. Of the three HP1 isoforms HP1γ expression is the highest in CMs and HP1γ is the only isoform showing clear co-localization with H3K9me3 in ACMs (Fig. 1b, c and Additional file 1: Fig. S1). This finding suggested that HP1γ could be the critical isoform in CMs. We generated a cardiac-specific HP1γ deletion mouse model, which is the first report of a tissue-specific HP1γ KO. However, deleting HP1γ had minimal effect on global gene expression and no significant effect on cell cycle gene silencing, cardiac growth or function. This may be related to a compensatory effect of other HP1 family members, particularly HP1β, because HP1β expression was upregulated in HP1γ KO mice and its localization to the perinuclear region, which is normally occupied by HP1γ, was enhanced.
All HP1s contain two very well conserved domains, the CD and CSD, which allows different HP1 family proteins to form heterodimers with each other. Although slight binding preferences might exist between the HP1 family members, HP1s share many binding partner proteins [19, 21, 37, 38, 43]. These findings indicate that HP1s have overlapping roles and redundant functions, which is supported by this study where cardiac-specific HP1γ KO has a very mild phenotype. We also found that significant upregulation and altered localization of HP1β in HP1γ KO CMs (Fig. 3a, c and Additional file 6: Fig. S4B), suggesting compensatory upregulation. Upregulation of other HP1 isoforms was not reported in HP1γ hypomorphic or gene trapped HP1γ KO cells [23, 34, 35]. These findings suggest that compensatory upregulation could be cell-type specific. In contrast to our finding, indispensable roles for HP1s was reported in mice where there was constitutive disruption of HP1γ expression. Germline KO or a hypomorphic allele of HP1γ mouse showed defects in spermatogenesis and primordial germ cells [23, 34, 35]. Similarly, germ line deletion of HP1β showed aberrant cerebral cortex development and is perinatal lethal . Thus, the necessity and role of each HP1 isoform appear to be cell type and context dependent. This study demonstrated that HP1γ is a dispensable, at least for normal cardiac growth and function.
HP1s have been shown to establish heterochromatin by mediating methylation of two of histone tails, H3K9me2/3 and H4K20me3. H3K9me2 and H3K9me3 are created by HMT G9a and Suv39h1, respectively [45, 46]. However, we did not detect any differences in global H3K9me2/3 level, indicating that HP1γ is dispensable to maintain H3K9me2/3 in CMs. In contrast, we found a significant reduction in H4K20me3 in HP1γ KO CM. Although both Suv420h1 and Suv420h2 can mediate H4K20me3 in vitro , Suv420h2 is the enzyme likely to establish H4K20me3 in vivo . Suv420h2 has been shown to interact with all isoform of HP1s in both mouse and human [38, 48] and it is believed that targeting of Suv420h2 to chromatin is HP1 dependent, at least onto pericentromeric heterochromatin . Recent work has shown that HP1β is likely to regulate targeting of Suv420h2 in mouse embryonic fibroblasts . By contrast, we find that HP1γ is critical in regulating H4K20me3 levels and distribution in CMs. These data indicate that the regulation of H4K20me3 is likely to be cell-type specific and dependent upon HP1-isotype-specific interactions with Suv420h2.
Previous studies have reported a potential role of H3K9me3 and HP1s on cardiac hypertrophy including regulation of fetal cardiac gene expression [40, 51]. However, we found no effect of HP1γ deletion on cardiac gene expression (ANP, βMHC and αMHC) at baseline or after TAC (Figs. 4c and 6d). CM size and cell cycling was not changed in HP1γ KO mice either (Fig. 4b, c and Additional files 11: Fig. S7 and 12: S8). Although there was slight increase in heart mass in TAC-induced model, there was no difference of heart mass in isoproterenol-induced hypertrophy model in HP1γ KO mice (Additional file 14: Fig S10). These data suggest that HP1γ has a dispensable role in mediating cardiac hypertrophy.
Recent studies have shown that HP1s have a gene regulatory role, not only in gene silencing, but also activation [52, 53]. To our surprise, loss of HP1γ showed very moderate effect on gene expression in CMs (Fig. 5a, Additional file 9: Fig. S6 and Fig. 7a, b). Our RNA-seq and qPCR confirmed that Cyp2b10 and Dlgap1 were upregulated in HP1γ KO CMs (Fig. 5b and Additional file 13: Fig. S9). These genes are poorly expressed in CMs and consistent with this fact, ENCODE data show that promoters of these genes have limited accessibility, lack of active histone mark (H3K9ac) and no recruitment of Polymerase II in the heart , which suggest that these genes are heterochromatic. It is reasonable to hypothesize that HP1γ is suppressing these genes by HP1’s classic heterochromatin recruiting mechanism.
We created a cardiac-specific HP1γ KO mouse model to test the role of HP1γ in cell cycle gene silencing and cardiac growth. To our knowledge, this is the first report examining the tissue- and isoform-specific role of HP1γ using a conditional KO system. We found that HP1γ has a significant role on H4K20me3 maintenance, although both HP1γ and H4K20me3 seem to be dispensable for cardiac cell cycle gene silencing and growth. Recent studies have reported the dispensability of classical heterochromatin regulators such as H3K9me2/3  and DNA methylation  on gene silencing in vivo model, even though significant roles of these genes have been determined in vitro [16, 22, 57, 58]. This study highlights the complexity of epigenetic regulation and importance of examining the function of these regulators in vivo.
KO designed and performed most of experiments as well as participated in interpretation, preparing figures and writing. DE, CF and HK performed experiments included in manuscript. JPB and PBS generated HP1γfl mouse. WRM conceived of the study, participated in its design and coordination and reviewed the manuscript. All authors read and approved the final manuscript.
We thank Yonggang Liu for performing TAC surgery.
The authors declare that they have no competing interests.
Availability of data and materials
The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Consent for publication
Ethics approval and consent to participate
This work is supported by American Heart Association (14POST20320016 awarded to KO) and NHLBI HL70748 to WRM. The work in the PBS was in part supported by the grant from the Ministry of Education and Science of Russian Federation #14.Y26.31.0024.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
- 6.Zhang R, Poustovoitov MV, Ye X, Santos HA, Chen W, Daganzo SM, Erzberger JP, Serebriiskii IG, Canutescu AA, Dunbrack RL, Pehrson JR, Berger JM, Kaufman PD, Adams PD. Formation of MacroH2A-containing senescence-associated heterochromatin foci and senescence driven by ASF1a and HIRA. Dev Cell. 2005;8:19–30.CrossRefPubMedGoogle Scholar
- 13.El-Nachef D, Oyama K, Wu Y-Y, Liu Yonggang, Zhang Y, MacLellan WR. Epigenetic control of adult cardiac myocyte proliferation [abstract]. Circulation. 2016;134:A15971.Google Scholar
- 14.Eissenberg JC, James TC, Foster-Hartnett DM, Hartnett T, Ngan V, Elgin SC. Mutation in a heterochromatin-specific chromosomal protein is associated with suppression of position-effect variegation in Drosophila melanogaster. Proc Natl Acad Sci USA. 1990;87:9923–7.CrossRefPubMedCentralPubMedGoogle Scholar
- 27.Weldy CS, Liu Y, Liggitt HD, Chin MT. In utero exposure to diesel exhaust air pollution promotes adverse intrauterine conditions, resulting in weight gain, altered blood pressure, and increased susceptibility to heart failure in adult mice. PLoS ONE. 2014;9:e88582.CrossRefPubMedCentralPubMedGoogle Scholar
- 29.Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, van Baren MJ, Salzberg SL, Wold BJ, Pachter L. Transcript assembly and quantification by RNA-Seq reveals unannotated transcripts and isoform switching during cell differentiation. Nat Biotechnol. 2010;28:511–5.CrossRefPubMedCentralPubMedGoogle Scholar
- 44.Aucott R, Bullwinkel J, Yu Y, Shi W, Billur M, Brown JP, Menzel U, Kioussis D, Wang G, Reisert I, Weimer J, Pandita RK, Sharma GG, Pandita TK, Fundele R, Singh PB. HP1-beta is required for development of the cerebral neocortex and neuromuscular junctions. J Cell Biol. 2008;183:597–606.CrossRefPubMedCentralPubMedGoogle Scholar
- 46.Tachibana M, Sugimoto K, Nozaki M, Ueda J, Ohta T, Ohki M, Fukuda M, Takeda N, Niida H, Kato H, Shinkai Y. G9a histone methyltransferase plays a dominant role in euchromatic histone H3 lysine 9 methylation and is essential for early embryogenesis. Genes Dev. 2002;16:1779–91.CrossRefPubMedCentralPubMedGoogle Scholar
- 47.Schotta G, Sengupta R, Kubicek S, Malin S, Kauer M, Callen E, Celeste A, Pagani M, Opravil S, De La Rosa-Velazquez IA, Espejo A, Bedford MT, Nussenzweig A, Busslinger M, Jenuwein T. A chromatin-wide transition to H4K20 monomethylation impairs genome integrity and programmed DNA rearrangements in the mouse. Genes Dev. 2008;22:2048–61.CrossRefPubMedCentralPubMedGoogle Scholar
- 50.Bosch-Presegue L, Raurell-Vila H, Thackray JK, Gonzalez J, Casal C, Kane-Goldsmith N, Vizoso M, Brown JP, Gomez A, Ausio J, Zimmermann T, Esteller M, Schotta G, Singh PB, Serrano L, Vaquero A. Mammalian HP1 isoforms have specific roles in heterochromatin structure and organization. Cell Rep. 2017;21:2048–57.CrossRefPubMedGoogle Scholar
- 55.Nuhrenberg TG, Hammann N, Schnick T, Preissl S, Witten A, Stoll M, Gilsbach R, Neumann FJ, Hein L. Cardiac Myocyte De Novo DNA Methyltransferases 3a/3b Are Dispensable for Cardiac Function and Remodeling after Chronic Pressure Overload in Mice. PLoS ONE. 2015;10:e0131019.CrossRefPubMedCentralPubMedGoogle 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.