Egr1 loss-of-function promotes beige adipocyte differentiation and activation specifically in inguinal subcutaneous white adipose tissue

In mice, exercise, cold exposure and fasting lead to the differentiation of inducible-brown adipocytes, called beige adipocytes, within white adipose tissue and have beneficial effects on fat burning and metabolism, through heat production. This browning process is associated with an increased expression of the key thermogenic mitochondrial uncoupling protein 1, Ucp1. Egr1 transcription factor has been described as a regulator of white and beige differentiation programs, and Egr1 depletion is associated with a spontaneous increase of subcutaneous white adipose tissue browning, in absence of external stimulation. Here, we demonstrate that Egr1 mutant mice exhibit a restrained Ucp1 expression specifically increased in subcutaneous fat, resulting in a metabolic shift to a more brown-like, oxidative metabolism, which was not observed in other fat depots. In addition, Egr1 is necessary and sufficient to promote white and alter beige adipocyte differentiation of mouse stem cells. These results suggest that modulation of Egr1 expression could represent a promising therapeutic strategy to increase energy expenditure and to restrain obesity-associated metabolic disorders.


Egr1 −/− mice display spontaneous increase of Ucp1 expression specifically in SC-WAT . Egr1
loss-of-function in postnatal and 4-month-old mice leads to a reduced body weight despite similar weight of SC-WAT fat pads in control Egr1 +/+ and mutant Egr1 −/− mice 20 . Significant body weight reduction was confirmed in 8-month-old Egr1 -/mice (Fig. 1A). In addition, all isolated white subcutaneous (SC-WAT), gonadic (GAT), mesenteric (MAT) and perirenal (PAT) and brown (BAT) fat pads display similar percentage of body weight in control Egr1 +/+ and mutant Egr1 -/mice (Fig. 1B, C). These observations indicate that lower body weight observed in mutant Egr1 -/mice is not due to a reduced adipose tissue mass, and suggest that Egr1 lossof-function reduces body weight through unknown global metabolic processes. Egr1 loss-of-function has been shown to induce spontaneous browning of SC-WAT, associated with increased Ucp1 expression, without external stimulation 20 . Here, we quantified the expression of Egr1, Cebpb, Pparg, Ucp1, Lep and Dcun1d3 in five different adipose tissue depots from control Egr1 +/+ and Egr1 −/− mice. As expected, Egr1 was nearly undetectable in mutant Egr1 -/fat pads ( Fig Fig. 1 and 2A-C) except in MAT in which Lep expression was lower in mutants. These observations are consistent with previous results showing the absence of spontaneous browning in BAT and GAT from Egr1-deleted mice 23 . Our results indicate that Egr1 deficiency induces a browning phenotype restricted to SC-WAT, and suggest that Egr1 is required to repress SC-WAT browning in standard housing conditions. SC-WAT browning in mutant Egr1 −/− mice is characterized by an increased number of beige adipocytes at the expense of white adipocytes 20 . Taken together, our results suggest that Egr1 represses Cebpb, Ucp1 and activates Lep expression in SC-WAT, which is in accordance with previous results showing recruitment of Egr1 to their promoters 20,25 . Decreased Lep expression in MAT depot in Egr1-null mice (MAT, Supplementary  Fig. 2B) suggests that Egr1 can activate Lep independently of the browning phenotype. At last, as previously observed, Pparg expression was not affected by Egr1 loss-of-function in SC-WAT 20  Dcun1d3 has been shown upregulated in mutant Egr1 -/-SC-WAT 20 . Dcun1d3 is expressed in adipocytes within adipose tissue (Fig. 2B), however its role has not been elucidated yet. While Dcun1d3 was present in all adipose depots in controls ( Fig. 2A, Supplementary Fig. 1 and 2A-C), its expression was specifically upregulated in Egr1-deficient SC-WAT ( Fig. 2A, Fig. 2A, Supplementary Figs. 1 and 2A-C). This suggests that Dcun1d3 may be related to SC-WAT browning in mutant Egr1 −/− mice.
Taken together, these results indicate that Egr1 loss-of-function increases Cebpb, Ucp1 and Dcun1d3 expression and decreases Lep expression in SC-WAT, resulting in spontaneous browning in this specific adipose tissue depot. One hypothesis could be that Egr1 interacts with specific co-regulators in fat pads, leading to the regulation of different genetic programs in each depot. The co-factors allowing Egr1 to regulate the browning program might be specifically expressed in SC-WAT.
Egr1 loss-of-function induces a metabolic shift in white adipose tissue, leading to a BAT-like oxidative metabolism. We then investigated whether the brown transcriptomic signature observed in mutant Egr1 −/− SC-WAT was correlated with metabolic modifications. While control Egr1 +/+ SC-WAT is characterized by a low mitochondrial mass, marked by a low mitochondrial/nuclear DNA ratio, we observed an increased mitochondrial mass in mutant SC-WAT (Fig. 3A), with a ratio being much similar to that of BAT. Such increase is specific to this tissue and was not observed in any other tested fat tissues ( Supplementary Fig. 3).

Scientific RepoRtS
| (2020) 10:15842 | https://doi.org/10.1038/s41598-020-72698-w www.nature.com/scientificreports/ To examine whether this increased mitochondrial mass has functional metabolic consequences, we assessed oxygen consumption rate (OCR) of BAT, SC-WAT and GAT freshly dissected from control and mutant mice. Using Seahorse XFe24 analyzer, we showed that while Egr1 loss-of-function does not affect GAT (Supplementary Fig. 3B, C) or BAT OCR (Fig. 3B, C, D), it leads to an increase of all markers of oxidative capacity, including basal and maximal OCR, ATP-linked OCR and non-mitochondrial OCR. Importantly, all rates quantified in mutant Egr1 −/− SC-WAT were close to that obtained with BAT, demonstrating a metabolic shift of Egr1 −/− SC-WAT towards a BAT-like oxidative metabolism. As expected, this increase of oxidative capacity does not lead to altered ROS production, as indicated by similar carbonylation protein levels ( Supplementary Fig. 3D), thus demonstrating that the browning process observed in Egr1 mutant occurs in physiological conditions. Our previous study has revealed a spontaneous browning of the SC-WAT in Egr1 mutant, with Ucp1 expression being up-regulated in absence of external stimulation 20 . We now show that this browning also exerts at physiological level with consequences on metabolic and bio-energetic activity of SC-WAT. Our results reveal a shift of Egr1 -/-SC-WAT towards a BAT-like oxidative metabolism and are correlated with previous studies establishing a link between increased Egr1 expression and obesity 22,23 . They suggest a central role of Egr1 in regulation of SC-WAT homeostasis in the adult.
Egr1 is necessary to promote white and to repress beige adipocyte differentiation of mouse adipose stem cells. To further analyze the role of Egr1 in spontaneous browning of SC-WAT in mutants, we tested whether Egr1 loss-of-function can alter the differentiation of SC-WAT-derived mouse adipose stem   Histogram represents the basal OCR (determined as the difference between OCR before oligomycin and OCR after rotenone/antimycin A), maximal OCR (difference between OCR after FCCP and OCR after rotenone/antimycin A), ATP-linked OCR (difference between OCR before and after oligomycin), and the non-mitochondrial OCR (OCR after rotenone and antimycin A treatment) calculated from data obtained in A. Error bars represent the means + standard deviations with n = 3 animals for each genotype, *p < 0.05.  (Fig. 4A). In contrast, only 20% of mutant mASC-Egr1 −/− cells exhibited lipid droplets (Fig. 4A) suggesting that Egr1 is necessary for white adipocyte differentiation. To confirm this hypothesis, we quantified the expression of adipocyte markers in control and mutant cells at different days after differentiation. As expected, Egr1 is highly expressed during the entire white adipocyte differentiation of control mASC-Egr1 +/+ cells, while not detected in mutant mASC-Egr1 −/− cells (Fig. 4B). Cebpb and Pparg encode transcription factors belonging to the core genetic program that trigger adipogenesis 26 . In both control mASC-Egr1 +/+ and mutant mASC-Egr1 −/− , Cebpb expression first decreases after 4 days of white differentiation and returns to basal level at the end of the process. Pparg expression continuously increases until day 8, and returns to basal level at the end of white differentiation. These results indicate that Egr1 is not necessary to activate Cebpb and Pparg expression during white adipocyte differentiation. Since Egr1 has been described as a direct regulator of Cebpb expression by chromatin immunoprecipitation assays 20 , our observations suggest that absence of Egr1 could be compensated in mASC-Egr1 -/cells to activate Cebpb expression, or that Egr1 is recruited to Cebpb promoter only as a transcriptional repressor of beige adipocyte differentiation 20 . The beige adipocyte marker Ucp1 was not expressed either in control or mutant cells. During white adipocyte differentiation, Lep, a late marker of white adipocyte differentiation 27,28 , was significantly down-regulated in mutant mASC-Egr1 -/cells compared to controls, which is consistent with the reduced white adipocyte differentiation observed in Egr1 −/− cells (Fig. 4A). Resistin (Retn) is an adipocyte-specific hormone secreted by WAT 3 . In control and mutant cells, Retn expression was absent at day 0, increased during white differentiation reaching a peak at day 4, and subsequently decreased during late differentiation (Fig. 4B). A similar Retn expression profile has been previously described during white adipocyte differentiation of 3T3-L1 preadipocytes 29 . This result indicates that Egr1 may not regulate Retn expression or that the absence of Egr1 was compensated. Dcun1d3 displays a similar expression profile to Retn during white differentiation of control mASC-Egr1 +/+ cells. In contrast to what we previously observed in vivo with the Egr1 −/− mutant ( Fig. 2A), Egr1 loss-of-function did not affect Dcun1d3 expression during mASC white adipocyte differentiation (Fig. 4B). This result can be explained by the difference between in vivo analysis of a mixed tissue, and study of an in vitro homogenous cell population strictly directed towards white differentiation.

Scientific RepoRtS
These observations demonstrate that while mutant mASC-Egr1 −/− cells express some markers at a similar level than in control cells (Cebpb, Pparg, Retn and Dcun1d3, Fig. 4B), they present significant morphological (Fig. 4A) and transcriptomic (Fig. 4B, Lep) alterations in late markers of normal white adipocyte differentiation. This suggests that Egr1 has a positive role on white adipocyte differentiation.
Following beige adipogenic stimulation, control mASC-Egr1 +/+ cells displayed a low level (17%) of differentiation into beige adipocytes marked by the appearance of numerous small lipid droplets in their cytoplasm (Fig. 4C). In contrast, Egr1 deletion led to a robust differentiation of 78% of mutant mASC-Egr1 −/− cells into beige adipocytes (Fig. 4C). This result shows that beige adipocyte differentiation in wild type mASC is a less efficient process compared to white differentiation (Fig. 4A, C). More importantly, this observation reveals that Egr1 acts as a critical repressor of beige adipocyte differentiation. In wild type mASC, Egr1 expression increases during the first time-points of beige adipocyte differentiation and decreases at the end of the process (Fig. 4C). In contrast to white adipocyte differentiation, Cebpb expression increases during wild type mASC beige adipocyte differentiation whereas Pparg expression remains constant from day 4 (Fig. 4D). Interestingly, Egr1 loss-of-function significantly increases Cebpb and Pparg expression, suggesting that Egr1 is necessary to repress Cebpb and Pparg expression during beige adipocyte differentiation. As expected, Ucp1 expression is significantly upregulated in mutant mACS-Egr1 −/− compared to control cells (Fig. 4D), consistently with the browning phenotype of Egr1 −/− mice (Figs. 2 and 3) 20 . Lep is not expressed during beige adipocyte differentiation of both control and or beige (D) differentiation medium, cells were used for RNA purification. RNA samples were analysed by RT-qPCR analysis for Egr1, Dcun1d3, Ucp1, Lep and Retn expression in control mASC-Egr1 +/+ and mutant mASC-Egr1 −/− cells. Transcripts are shown relative to the level of Rplp0 and 18S transcripts. The relative mRNA levels were calculated using the 2 −∆∆Ct method, with control being normalized to 1. For cells subjected to white differentiation, experiments were performed with n ≥ 5 independent cultures for each genotype (n = 12 at Day0, n = 6 at Day4, n = 5 at Day8 and n = 10 at Day12 for control mASC-Egr1 +/+ cells; n = 11 at Day0, n = 6 at Day4, n = 6 at Day8 and n = 12 at Day12 for mutant mASC-Egr1 −/− cells). For cells subjected to beige differentiation, experiments were performed with n = 16 independent cultures for each timepoint and genotype. Error bars represent the mean + standard deviations. The p values were obtained using the Mann-Withney test. Asterisks * indicate the p values of gene expression levels in mASC-Egr1 +/+ and mASC-Egr1 −/− compared to the first day of gene detection *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001; # indicate the p values of gene expression levels in mASC-Egr1 +/+ versus mASC-Egr1 −/− , for each time point, # P < 0.05, ## P < 0.01, ### P < 0.001, #### P < 0.0001. www.nature.com/scientificreports/ mutant mASC (Fig. 4D). Expression of Retn is detected during beige adipocyte differentiation (Fig. 4D), while at lower levels compared to white differentiation, which further suggests that Retn may be a generic marker of both white and beige adipocyte differentiation 13 . Similarly, mutant mASC-Egr1 −/− cells which exhibit a robust differentiation rate, visualized by accumulation of lipid droplets in the cytoplasm (78%, Fig. 4C), are characterized by higher Retn expression levels compared to wild type cells (Fig. 4D). This further confirms that Retn expression can be associated with beige adipocyte differentiation. Dcun1d3 expression is increased during beige adipocyte differentiation (Fig. 4D). In addition, Egr1 loss-of-function leads to a significant upregulation in Dcun1d3 expression (Fig. 4D), suggesting that Egr1 represses Dcun1d3 expression. This is similar to our in vivo results (Fig. 2B). Altogether, our results indicate that Egr1 is necessary to promote white adipocyte differentiation and to repress beige adipocyte differentiation of mASC. Consistent with the cell differentiation phenotypes, Egr1 is necessary to activate Lep during white differentiation and to repress Ucp1 during beige adipocyte differentiation. Retn and Dcun1d3 are expressed during both white and beige adipocyte differentiation of mASC, suggesting a generic role of these factors during adipocyte differentiation (Fig. 7). Interestingly, Egr1 loss-of-function specifically affects the expression of the generic adipocyte differentiation markers Cebpb, Pparg, Retn and Dcun1d3 during beige adipocyte differentiation, which confirms the negative role of Egr1 on beige differentiation suggested by in vivo data (Figs. 2 and 3). The positive role of Egr1 on in vitro white adipocyte differentiation is consistent with our in vivo results (Figs. 2 and 3) and with studies correlating Egr1 overexpression with obesity 22,23 . Egr1 is sufficient to promote white and to repress beige adipocyte differentiation of C3H10T1/2 mesenchymal stem cells. To further characterize the role of Egr1 during adipocyte differentiation and confirm the results obtained with mASC, we used a second in vitro approach, taking advantage of the Tol2 genomic system to perform Egr1 gain-of-function experiments in C3H10T1/2 mesenchymal cells. The Tol2 system was used in combination with the viral T2A system, which allowed stable and bi-cistronic expression of two proteins 30 (Supplementary Fig. 4). Control C3H-Tom-Gfp cells co-expressed cytoplasmic Tomato and nuclear H2B-GFP fluorescent reporter proteins, whereas in C3H-Tom-Egr1 cells, the H2B-Gfp coding sequence was replaced with Egr1 coding sequence ( Supplementary Fig. 4). C3H-Tom-Gfp and C3H-Tom-Egr1 cells exhibited cytoplasmic Tomato expression and nuclear H2B-Gfp or Egr1 expression, respectively (Figs. 5A and 6A).
At the end of white adipogenic stimulation, 90% of C3H-Tom-Gfp and 95% of C3H-Tom-Egr1 cells differentiated into white adipocytes, marked by the appearance of large lipid droplets in the cytoplasm (Fig. 5B, C, white arrowheads). These observations indicate that Egr1 gain-of-function promotes C3H10T1/2 cell differentiation into white adipocytes and support previous findings showing increased Egr1 expression in WAT of obese mice and humans compared to lean individuals 22,23 . In contrast, while 97% of C3H-Tom-Gfp differentiated into beige adipocytes, indicated by the presence of small lipid droplets (Fig. 6B, C, white arrowheads), only 50% of C3H-Tom-Egr1 cells achieved beige differentiation. These results confirm that Egr1 gain-of-function in C3H10T1/2 mesenchymal stem cells reduces beige adipocyte differentiation 20 .
Ucp1 was not detected at the end of white differentiation (Fig. 5D). As expected, the expression of the beige markers Ucp1, Cidea and Cox8b was upregulated in control C3H-Tom-Gfp cells upon differentiation into beige adipocytes (Fig. 6D). A moderate decrease in Ucp1 expression was also observed in C3H-Tom-Egr1 cells when compared to control cells (Fig. 6D), in accordance with their impaired beige differentiation. Surprinsingly, an increase of both Cidea and Cox8b expression was also observed during white adipocyte differentiation (Fig. 5D). In addition, Egr1 gain of-function increases both Cidea and Cox8b expression during white (Fig. 5D) or beige (Fig. 6D) adipocyte differentiation. Egr1 therefore acts as an activator of Cidea and Cox8b expression during white and beige adipocyte differentiation of C3H10T1/2 and is able to repress Ucp1 expression only. One hypothesis could be that, in contrast to the in vivo context (Fig. 2) the co-factors allowing Egr1 to repress Cox8b and Cidea might not be expressed by C3H10T1/2 cells.
Both Retn and Dcun1d3 expression were upregulated in C3H-Tom-Gfp cells after white and beige differentiation (Figs. 5D and 6D), which strengthens the hypothesis of a generic role of these two genes during adipocyte differentiation. In addition, Dcun1d3 expression was not altered in C3H-Tom-Egr1 cells compared to controls (Figs. 5D and 6D), showing that Egr1 is not required to regulate Dcun1d3 expression during adipocyte differentiation. We could not detect any expression of the white adipocyte marker Lep either in C3H-Tom-Gfp or in C3H-Tom-Egr1 cells during white adipocyte differentiation. These results highlights the limitations of in vitro cell lines in the analysis of white adipocyte differentiation and confirms that primary mASC are the only in vitro system that allows Lep detection 31 .
Collectively, these results indicate that Egr1 is sufficient to promote white adipocyte differentiation of C3H10T1/2 cells and to impair beige differentiation, presumably by down-regulation of Ucp1 expression.
In summary, Egr1 deletion induces spontaneous WAT browning specifically in the inguinal subcutaneous depot. This process acts through Ucp1 upregulation and results in OCR increase at a level close to the BAT (Fig. 7). In vitro differentiation of mASC isolated from Egr1 mutant SC-WAT confirmed that Egr1 loss-offunction promotes beige adipocyte differentiation through Ucp1 upregulation, and repress white adipocyte www.nature.com/scientificreports/ differentiation. We confirmed these results using Egr1 gain-of-function experiments in C3H10T1/2 mesenchymal stem cells. Altogether, our study reveals that Egr1 is necessary and sufficient to promote full white adipocyte differentiation and sufficient to prevent beige adipocyte differentiation. These results are in accordance with the increased expression of Egr1 observed in obese mice and humans 22,23 and establishes Egr1 as a putative promising target to counteract obesity.  In situ hybridization to adipose tissue sections. The mouse Dcun1d3 fragment was amplified by PCR (reverse 5′ggtcatcactagccatcctaac; forward 5′cagtattcagggagggactttg) from SC-WAT cDNA and cloned in the PGEM-T easy plasmid (PROMEGA). The antisense probe was synthesized using T7 polymerase after SalI digestion. The sense probe was obtained using Sp6 polymerase after SacII digestion. Inguinal subcutaneous fat pads were isolated from 1-month-old female mice, fixed in 4% paraformaldehyde overnight and cryo-sectioned. 8 µm wax tissue sections were used for in situ hybridization as previously described 33 .

Methods
Quantification of mitochondrial DNA. WAT and BAT were dissected from 8-months-old control Egr1 +/+ and mutant Egr1 -/female mice and immediately frozen at −80 °C. DNA purification was performed using the QIAmp DNA Minikit (QIAGEN). DNA concentration and purity were measured using a Nanodrop. Quantitative real time PCR was performed on a StepOnePlus PCR machine (APPLIED BIOSYSTEMS) using the FastStart Universal SYBR Green Master (ROCHE) with primers that amplify the mitochondrial CytB gene and the nuclear encoded NDUFV1 gene as endogenous reference, as described in 34 .
Metabolic and bio-energetic analysis. Tissue bioenergetics were analyzed on a Seahorse extracellular flux analyzer (XF e 24) according to the manufacturer's instructions. 2 mg explants of WAT and BAT freshly dissected from 8-months-old control Egr1 +/+ and mutant Egr1 −/− female mice were immediately placed in a Seahorse XF e 24 islet plate in 500 µL of pre-heated minimal assay medium (SEAHORSE BIOSCIENCES, 37 °C, pH 7.40) supplemented with glucose (2.25 g/L), sodium pyruvate (1 mM) and L-glutamine (2 mM). The tissue weight was based on initial assays that optimized the oxygen consumption rate (OCR). After one hour of equilibration at 37 °C in a non-CO 2 incubator, OCR was measured successively in control condition (basal rates) and after successive injections of oligomycin (1 μM) that inhibits ATP synthase (ATP-linked OCR), FCCP (10 μM) that uncouples mitochondrial OXPHOS (maximal OCR) and mixed rotenone/antimycin (10 μM) that inhibit complexes I and III (non-mitochondrial OCR).
Detection and quantification of carbonylated proteins. WAT  After three washes using a solution 7% acetic acid and 10% ethanol, CF555DI hydrazide labeling was detected using a ChemiDoc MP Imagin System (BIORAD). Total protein staining was performed using Instant-Blue Protein Stain (EXPEDEON). Raw data were analyzed using the Fiji/ImageJ software. In both CF555DI and Coomassie's analysis, the protein profiles were quantified by measuring the intensity of the lane for each sample. Data were normalized using the CF555DI/Coomassie ratio.  INVITROGEN) supplemented with 10% foetal bovine serum (FBS, Sigma), 1% penicillin-streptomycin (SIGMA), 1% Glutamine (SIGMA), 800 μg/ml G418 Geneticin (SIGMA) and incubated at 37 °C in humidified atmosphere with 5% CO 2 as described in 20 . Cells were cultured in white adipocyte differentiation medium (STEMPRO, THERMOFISHER) until differentiation which takes 7 to 12 days. The medium was changed every 3-4 days. Day 0 corresponds to the addition of white differentiation medium, at 100% confluence. mASC-Egr1 +/+ and mASC-Egr1 -/subjected to white adipocyte differentiation medium were stopped at Day 0, Day 4, Day 8 and Day 10 or 12 for gene expression analysis and stopped at Day 10 for Oil Red O staining (as described in previous study 20 . C3H-Tom-Gfp and C3H-Tom-Egr1 cells subjected to white adipocyte differentiation medium were stopped at Day 0 and at the end of differentiation for gene expression analysis. Confluent cells were cultured in beige differentiation induction medium for 2 days and in beige maturation medium for 7 to 10 days according to published protocols 20,38 . Day 0 corresponds to the addition of beige differentiation induction medium at confluence. The maturation medium was changed every 3-4 days. mASC-Egr1 +/+ and mASC-Egr1 -/subjected to beige adipocyte differentiation medium were fixed for Oil Red O staining at Day 10 or lysed for gene expression analysis at Day 0, Day 4, Day 8 and Day 10.

Isolation of murine adipose stem/stromal cells (mASC
C3H-Tom-Gfp and C3H-Tom-Egr1 cells subjected to beige adipocyte differentiation medium were stopped at Day 0 and at the end of differentiation for gene expression analysis.
Cell number measurements were performed using the free software Fiji 39 .
Reverse-transcription and quantitative real time PCR. Total RNAs extracted from mouse fat pads, mASC-Egr1 +/+ and mASC-Egr1 -/-, or C3H-Tom-Gfp and C3H-Tom-Egr1 cells were Reverse Transcribed using the High Capacity Retro-transcription kit (Applied Biosystems). Quantitative PCR analyses were performed using SYBR Green PCR Master Mix (Applied Biosystems) and primers listed in Supplementary Table 1. The relative mRNA levels were calculated using the 2 −∆∆Ct method 40  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://creat iveco mmons .org/licen ses/by/4.0/.