Iron‐triggered signaling via ETS1 and the p38/JNK MAPK pathway regulates Bmp6 expression

BMP6 is an iron‐sensing cytokine whose transcription in liver sinusoidal endothelial cells (LSECs) is enhanced by high iron levels, a step that precedes the induction of the iron‐regulatory hormone hepcidin. While several reports suggested a cell‐autonomous induction of Bmp6 by iron‐triggered signals, likely via sensing of oxidative stress by the transcription factor NRF2, other studies proposed the dominant role of a paracrine yet unidentified signal released by iron‐loaded hepatocytes. To further explore the mechanisms of Bmp6 transcriptional regulation, we used female mice aged 10–11 months, which are characterized by hepatocytic but not LSEC iron accumulation, and no evidence of systemic iron overload. We found that LSECs of aged mice exhibit increased Bmp6 mRNA levels as compared to young controls, but do not show a transcriptional signature characteristic of activated NFR2‐mediated signaling in FACS‐sorted LSECs. We further observed that primary murine LSECs derived from both wild‐type and NRF2 knock‐out mice induce Bmp6 expression in response to iron exposure. By analyzing transcriptomic data of FACS‐sorted LSECs from aged versus young mice, as well as early after iron citrate injections, we identified ETS1 as a candidate transcription factor involved in Bmp6 transcriptional regulation. By performing siRNA‐mediated knockdown, small‐molecule treatments, and chromatin immunoprecipitation in primary LSECs, we show that Bmp6 transcription is regulated by iron via ETS1 and p38/JNK MAP kinase‐mediated signaling, at least in part independently of NRF2. Thereby, these findings identify the new components of LSEC iron sensing machinery broadly associated with cellular stress responses.

that mice with endothelial-specific Bmp6 deletion recapitulate phenotypes of the systemic Bmp6 knock-out (KO) model. 8e mechanistic details of LSEC Bmp6 regulation in response to iron-triggered cues are the focus of an intense investigation.One model proposes cell-autonomous regulation of Bmp6 by iron deposition in LSECs.Markers of LSEC iron accumulation in wild-type (WT) mice fed an iron-rich diet were shown to correlate with Bmp6 mRNA induction. 7Likewise, hemochromatotic ZIP14 knock-out mice, characterized by iron accumulation primarily in liver non-parenchymal cells but not in hepatocytes, showed elevated Bmp6 mRNA levels. 9[12] Mechanistically, Bmp6 was proposed to be a target of the transcription factor NRF2, a key sensor of oxidative stress which arises upon iron excess.However, direct and cell-autonomous involvement of NRF2 in iron-triggered Bmp6 induction specifically in LSECs was not formally demonstrated. 11 the contrary, some reports implied a non-cell-autonomous control of Bmp6 by iron cues.Changes in Bmp6 mRNA levels in LSEC isolated from Hjv knock-out mice, a model of severe hemochromatosis, and Tmprss6 knock-out animals, a model of iron-refractory iron deficiency anemia, reflect hepatocytic, not endothelial iron content. 7nsistent with these observations, a recent study demonstrated that primary LSECs induce Bmp6 upon iron exposure only in the presence of hepatocyte-derived protein factor and fail to increase Bmp6 transcription upon pharmacological NRF2 activation. 13Taken together, the above observations suggest that the precise signaling mechanisms governing the Bmp6 transcriptional response to iron-rich conditions remain not completely understood.
Here, we employed a murine model of aging that exhibits mild and physiological iron overload in the liver to investigate the mechanisms of Bmp6 regulation.By analyzing the transcriptomes of aged LSECs, as well as LSECs isolated from iron-injected mice, complemented by studies in primary murine LSECs, we show that Bmp6 transcription is regulated in an NRF2-independent manner via the transcription factor ETS1.We further provide evidence that the activities of the stress-inducible MAP kinases JNK and p38, the latter likely acting upstream of ETS1, are required for iron-dependent Bmp6 regulation.

| Mice and iron citrate (FeCit) injections
The aging experiments were conducted using female WT C57BL/6J mice, as reported previously 14 or NRF2 KO (B6.129X1-Nfe2l2 tm1Ywk /J) female and male mice.Aged-matched male WT C57BL/6J and NRF2 KO mice or C57BL/6J WT females (8-12 weeks old) were used for the generation of primary liver cell cultures.Female BALB/c mice (8-10 weeks old) were intravenously injected with a sterile solution of 150 μg of FeCit (Sigma-Aldrich) and citrate buffer (0.05 M) for 5 h.The injection procedure was approved by the local ethical committee in Warsaw (WAW2/152/2020).All animal experiments were conducted following the guidelines of the EU legislation and FELASA.Details are included in the Data S1.

| Liver primary cell culture and cell treatments
Liver cells were isolated from 8 to 10 weeks-old female C57BL/6J mice according to the standard two-step perfusion method 15 with minor modifications, and subjected to treatments as described in detail in the Data S1.

| ETS1 siRNA-mediated knockdown
Cells were transfected with siRNA targeting ETS1 or non-targeting siRNA using Lipofectamine RNAiMAX (Invitrogen) according to the manufacturer's instructions and treated with iron 48 h later, as described in the Data S1.

| Liver sinusoidal endothelial cells fluorescence-activated cell sorting (FACS) and flow cytometry analysis
Details are reported in the Data S1.

| Western blot
The detailed protocol is provided in the Data S1.

| Tissue and serum iron quantification
Tissue non-heme iron content was measured using the bathophenanthroline method. 16Transferrin saturation was determined with serum iron content (SFBC) and unsaturated iron-binding capacity kits (UIBC) (Biolabo).Total iron content in liver cells was measured using Iron Assay Kit (Sigma-Aldrich) according to the manufacturer's protocol and normalized to cell numbers.

| DAB-enhanced Perls staining
DAB-enhanced Perls staining was performed according to a standard method as described in the Data S1.

| RNA isolation and qPCR
RNA from tissue/sorted cells was extracted using TRIzol™ (LS) Reagent (ThermoScientific) according to the manufacturer's instructions, involving RNA precipitation with glycogen (ThermoScientific) for sorted cells.Details for RNA reverse transcription and real-time qPCR are described in the Data S1.

| RNA-Seq analysis
Transcriptome analysis of LSEC cells was conducted using Ion AmpliSeq™ Transcriptome Mouse Gene Expression Kit (Thermo Scientific) following the manufacturer's recommendations.Targeted cDNA fragments were amplified using the Ion AmpliSeq™ Library Kit 2.0 and measured using Agilent™ 2100 Bioanalyzer™.Further details, including methods for data analysis, are included in the Data S1.Data are deposited in the GEO repository under the accession number GSE235976.

| Statistical analysis
Statistical analysis was conducted using GraphPad Prism.Two-tailed unpaired Welch's t-test was applied for comparisons between two groups, and one-or two-way ANOVA with post-hoc Tukey's test was used for multiple comparisons.Changes between means were considered significant for p < .05.

| Liver iron accumulation during aging increases LSEC Bmp6 expression independently of oxidative stress and NRF2-mediated signaling
0][11] However, little is known about Bmp6 regulation in the context of non-pathological states.
Aging is characterized by tissue iron deposition, 14 but interestingly also hallmarked by a decline in NRF2-mediated signaling. 19To investigate the regulation of Bmp6 mRNA expression under such conditions, we compared 8-to 10-week-old young females with 10-11-monthold mice.We confirmed that aged mice show liver iron deposition in comparison to young controls (Figure 1A), accompanied by a drop in transferrin saturation (Figure 1B), thus representing a model of hepatic iron accumulation without the characteristics of systemic iron overload.Interestingly, we found that the total iron content of purified hepatocytes increased significantly during aging, whereas it remained unchanged at a relatively low level in nonparenchymal cells (NPCs) (Figure 1C).Likewise, DAB-enhanced Perls staining indicated hepatocytic iron deposition in aged versus young livers (Figure 1D) and we found that L-ferritin was elevated in aged hepatocytes, but not in LSECs (Figure 1E,F; see Figure S1 for representative LSEC gating strategy for flow cytometry).Consistently, we observed a trend for a mild increase in lipid peroxidation in aged livers (Figure S2A) and an induction of an oxidative stress marker Nqo1 (Figure 1G) but not of Hmox1 or Gclc (Figure S2B,C), while LSECs did not exhibit increased ROS levels or altered FPN expression (Figure 1H,I).Despite the lack of oxidative stress markers in LSECs, we found that the expression levels of Bmp6 and hepcidin were elevated in the aged compared to young livers (Figure 1J,K).Likewise, gene expression signatures of FACS-sorted LSECs revealed significant induction of Bmp6 (Figure 1L), but not of Hmox1, Nqo1, and Gclc which are considered NRF2 target genes 11 (Figure 1M-O; see Figure S1 for representative LSEC purity after sorting).In addition, we found that FACS-sorted LSECs from aged NRF2 knock-out (KO) mice exhibited increased Bmp6 mRNA levels compared to young controls, but other NRF2 target genes remained unaltered in total liver (Figure 1P and Figure S2D-F).Taken together, these data suggest that the regulation of Bmp6 in LSECs during aging is not associated with the activation of NRF2 signaling by ROS.

| Non-transferrin-bound iron-triggered induction of Bmp6 in primary LSECs requires oxidative stress but not NRF2
To further study iron-dependent Bmp6 regulation, we established an in vitro system of primary murine LSECs.Primary mouse liver NPC cultures were treated with NTBI in the form of iron citrate (FeCit), and the population of LSECs was subsequently FACS-sorted for gene expression analysis or analyzed with flow cytometry.We observed an upregulation of both Hmox1 and Bmp6 in LSECs in response to iron stimulation, with the highest level of induction observed at 18 h posttreatment (Figure 2A,B).We next sought to determine whether inhibition of iron-induced oxidative stress abrogates the Bmp6 response, a question that has not been yet addressed in primary LSECs.To this end, we employed two different antioxidants: Trolox, a vitamin E derivative, and MitoTEMPO, which specifically diminishes oxidative stress within the mitochondria.By using the CellROX fluorescent probe and measuring Hmox1 gene transcription in the sorted LSECs, we confirmed that both antioxidants were effective in mitigating iron-induced oxidative stress (Figure 2C,D).Importantly, we observed that quenching of oxidative stress reduced Bmp6 induction levels upon iron exposure (Figure 2E).Since oxidative stress was proposed to activate Bmp6 via NRF2, 11 we investigated this mechanism in our primary LSEC model.Interestingly, pharmacological stimulation of NRF2 by its activator CDDO-Im only slightly affected an early Bmp6 transcriptional response and failed to induce it at a longer time point (Figure 2F), despite a pronounced induction of the Hmox1 gene (Figure 2G).We ENCODE ChIP-seq data. 20,21In sum, these results indicate that the activation of Bmp6 transcription is triggered by the iron-dependent oxidative stress at least partly independently of NRF2, implying that other stress-responsive mechanisms may substitute for NRF2 functions.

| Analysis of aged LSEC transcriptome identified ETS1 as a mediator of iron-triggered Bmp6 induction
We hypothesized that still elusive mechanisms of NRF2-independent Bmp6 regulation may be shared between the response of the LSEC to iron loading and aging-related liver iron deposition.To gain an insight into such new possible modes of Bmp6 transcriptional control, we performed RNA-Seq analysis of LSECs FACS-sorted from aged versus young mice.We identified a relatively small number of differentially expressed genes (49 up-and 44 down-regulated; Figure 3A) that did not show functional enrichments.Next, we queried the set of upregulated genes into Cscan, a software that utilizes genome-wide ChIP-seq data to identify putative common regulators of input gene loci. 22Interestingly, we found that the hits were enriched in targets of ETS1 (Figure 3B), a transcription factor that binds proximal [close to the transcription start site (TSS)] and distal (approximately 2.2 kbp upstream of the TSS) regions within the promoter region of Bmp6 according to ENCODE data (Figure S3A).
Among the RNA-Seq hits, we validated another ETS1 target gene Ptgis identified by ChIP-seq data (Figure S3B), which was induced in aged LSECs (Figure 3C).We also detected decreased mRNA expression levels of Dusp6 (Figure 3D), consistent with the previously reported finding that Dusp6 is negatively regulated by ETS1. 23A similar pattern of Ptgis and Dusp6 regulation was observed in aged versus young NRF2 KO mice (Figure S4A,B).These analyses proposed ETS1 as an alternative NRF2-independent activator of Bmp6 expression, likely activated by both LSEC aging and iron accumulation.
We thus sought to verify the role of ETS1 in the iron-dependent regulation of Bmp6 expression in cultured primary LSECs.First, we found that the knockdown of ETS1 with two independent siRNAs, although not complete (Figure S5), diminished iron-dependent activation of the Bmp6 gene transcription (Figure 3E).We next sought to determine whether the pattern of gene regulation observed in aged LSECs is also present in primary LSECs treated with iron citrate.We observed that the positive and negative ETS1 target genes Ptgis and Dusp6, respectively, were regulated in response to iron exposure in cultured primary LSECs, independently of the presence of NRF2 (Figure 3F,G).
Finally, we aimed to examine the involvement of ETS1 in the early response of Bmp6 to iron injection in vivo.We found that intravenous injection of a relatively low dose of iron citrate elicited a marked induction of Bmp6 (Figure 3H) and other oxidative stress marker genes without causing a profound LSEC iron deposition (Figure S6).
Iron exposure provoked a pronounced response of the LSEC transcriptome.This was hallmarked by the upregulation of 254 genes, including not only redox-associated stress-inducible genes but also the endothelial growth factor Vegfd, the iron-containing transcriptional activator Pir, the ferroportin-encoding gene Slc40a1 and Bmp6 among the top 20 hits, and the downregulation of 47 transcripts (Figure 3I).Strikingly, the analysis of the induced genes using Cscan software identified ETS1 as the second top transcription factor mediating the gene expression signature in iron citrate-exposed LSECs (Figure 3J).Even though only seven genes, Aox1, Cdkn1a, Stc1, Zbtb38, Epb41, Fuca2, and Tcim, were significantly induced in transcriptomic data of both aged and iron citrate-exposed LSEC, most of them bind ETS1 according to ENCODE data.Among them, Aox1, a previously reported ETS1 target gene 24 was validated by secondary qPCR analysis in both scenarios (Figure 3K,L).Altogether, the combination of in vivo and in vitro studies proposed ETS1 as a transcription factor involved in Bmp6 regulation under iron-rich conditions, and likely beyond iron-triggered oxidative stress.

| Induction of Bmp6 gene expression by iron is not dependent on ERK1/2 but requires c-Jun N-terminal kinases and p38 mitogen-activated protein kinases activity
Next, we aimed to explore the mechanism underlying the activation of the ETS1 transcription factor by iron.Transcription factors with an ETS domain are activated by the mitogen-activated protein (MAP) kinases, 25,26 signaling cascades known to mediate cellular responses F I G U R E 2 Iron-triggered induction of Bmp6 in primary liver sinusoidal endothelial cells requires oxidative stress but not NRF2.(A and B) Liver non-parenchymal cells (NPCs) isolated from female C57BL/6J mice were treated with iron (FeCit, 1 mM) for 18 or 24 h.Hmox1 and Bmp6 mRNA expression levels were determined in FACS-sorted LSECs.(C-E) NPCs were treated with the antioxidant Trolox (20 μM) or the mitochondrial oxidative stress inhibitor MitoTEMPO (500 μM) 1 h before iron (FeCit, 1 mM) administration for 18 h.(C) Oxidative stress in the LSEC population was measured with a CellROX fluorescent probe using flow cytometry and mRNA expression levels of (D) Hmox1 and (E) Bmp6 were determined in FACS-sorted LSECs.(F and G) NPCs were exposed to the pharmacological NRF2 activator CDDO-Im (10 nM) for 6 or 18 h and mRNA expression levels of (F) Bmp6 and (G) Hmox1 were determined in FACS-sorted LSECs.(H-J) NPCs isolated from male C57BL/6J wild-type (WT) and NRF2 knock-out (NRF2 KO) mice were treated with 1 mM FeCit for 18 h, and the induction of (H) Bmp6, (I) Hmox1 and (J) Nqo1 mRNA levels were quantified in FACS-sorted LSECs.Rpl19 served as a reference gene for gene expression analysis.Data are expressed as mean ± SEM and each data point represents one biological replicate.Welch's unpaired t-test was used to determine statistical significance in panels A, B, F, and G, while two-way ANOVA with Tukey's Multiple Comparison test was used for panels C-E and H-J; ns-not significant, *p < .05,**p < .01,***p < .001,and ****p < .0001.[Color figure can be viewed at wileyonlinelibrary.com] to oxidative stress. 27MAP kinases are divided into three well-defined families: extracellular signal-regulated kinase (ERK), c-Jun N-terminal kinases (JNK), and p38 MAPK signaling. 28Thus, we decided to investigate the involvement of each of these pathways in mediating Bmp6 induction upon iron exposure.
Western blot results showed that the iron challenge did not induce ERK1/2 phosphorylation (Figure S7A).In parallel, we examined whether inhibition of ERK1/2 activation by the upstream kinase MEK1 suppresses the iron-dependent induction of Bmp6 transcription.Of note, we observed that the widely used inhibitor of MEK1-ERK1/2 signaling U0126 potently attenuated the levels of iron-triggered ROS, as previously reported 29 (Figure S7B).This off-target effect of U0126 was likely responsible for diminishing iron-induced Bmp6 expression (Figure S7C).By contrast, PD0325901, an inhibitor with no impact on iron-induced oxidative stress (Figure S7D) failed to suppress irontriggered transcriptional activation of Bmp6 (Figure S7E).Overall, these results suggest that the induction of Bmp6 transcription by iron is not attributable to the activation of the ERK1/2 kinase-dependent signaling pathway.
Using the same approaches, we further probed the involvement of the JNK and p38 pathways in iron-dependent Bmp6 regulation.We found that while the excess of iron had no statistically significant effect on the level of JNK kinase phosphorylation (Figure 4A), it enhanced the phosphorylation of p38 MAPK (Figure 4B).Interestingly, we found that specific inhibitors of both JNK (SP600125) and p38 (SB202190116) MAP kinases significantly reduced the ironinduced activation of Bmp6 transcription in FACS-sorted LSECs, without decreasing the ROS levels (Figure 4C,D).We further observed that the JNK inhibitor failed to abrogate the induction of the oxidative stress marker Hmox1 and the ETS1 target gene Ptgis (Figure 4E,F).
Conversely, the suppression of p38 activity with SB202190116 diminished the iron-triggered activation of both Hmox1 and Ptgis expression (Figure 4E,F).This evidence suggested that both branches of the MAPK signaling pathway, JNK and p38, participate in the transcriptional response of Bmp6 to iron loading, but only p38 likely acts upstream of ETS1.Of note, we also validated the MAPK p38-mediated Bmp6 and Ptgis induction using another common iron source, ferric ammonium citrate (FAC; Figure S8).
Finally, to obtain mechanistic insights into iron-dependent events at the Bmp6 promoter, we performed ChIP of RNA polymerase II (PolII), phospho-ETS1, and phospho-p38 MAPK, the latter with known chromatin binding functions, 30 against unspecific IgG.ChIP analysis showed basal PolII presence at the proximal promoter region of Bmp6 (prox) and the regulatory region located 2.2 kbp upstream from TSS (up) (Figure S3A), but only its recruitment close to TSS was enhanced by iron excess and diminished by p38 kinase inhibitor and antioxidants (Figure 4G).Strikingly both phospho-ETS1 and phospho-p38 MAPK were detected at the proximal and distal (À2.2 kbp) regions, their binding was significantly enhanced by iron treatment, and the interactions close to the TSS were diminished by p38 MAPK inhibitor and antioxidants (Figure 4H,I).None of the above observations were reflected at control beta-Actin (Actb) and Rhodopsin (Rho) genes that served as constitutively active and repressed loci, respectively (Figure S9).Taken together, our analysis suggests that the irontriggered transcriptional induction of the Bmp6 gene is dynamically regulated by ROS-dependent recruitment of activated phosphorylated ETS1 and MAPK p38 to chromatin at the Bmp6 promoter locus.
3.5 | LSECs exhibit a stronger induction of the Bmp6 in co-culture with hepatocytes in p38 MAPK-dependent manner Previous reports 7,13 and our aging model suggested that the regulation of Bmp6 transcription in LSECs may be affected by hepatocyte iron levels.Although in our hands primary LSECs induced Bmp6 expression upon iron exposure in a cell-autonomous manner, we aimed to verify if this response could be modified by the presence of hepatocytes.We observed that LSECs cultured in the presence of hepatocytes activated Bmp6 mRNA transcription to a greater extent upon iron supplementation compared to cells cultured without hepatocytes (Figure S10A).Interestingly, the increase in iron responsiveness in LSECs cultured with hepatocytes was coupled with an enhanced Hmox1 transcriptional response and a stronger induction of cellular ROS levels (Figure S10B,C).To better understand the complex nature of hepatocyte-LSEC cross-talk in Bmp6 regulation, we assessed whether it is associated with hepatocyte-dependent alterations of LSEC iron status.We found that H ferritin levels dropped substantially under the co-culture setting, and seemed to be enhanced upon iron challenge (albeit not significantly) only when hepatocytes were present (Figure S10D).The FPN levels were also induced to a higher extent in the co-culture settings, likely reflecting higher ROS levels (Figure S10E).We thus propose that low H ferritin levels in LSECs cultured with hepatocytes render LSECs more sensitive to ROS buildup.Of note, this phenomenon refers mostly to cytosolic ROS levels since the generation of lipid peroxides or mitochondrial ROS remained similar upon iron challenge in mono-and co-culture settings (Figure S10F,G).We further hypothesized that the synergy between iron-driven and hepatocyte-imposed signals in Identification of the transcription factor ETS1 as an activator of Bmp6 expression.(A) Shown is a volcano plot of differentially regulated genes identified by the RNA-Seq transcriptomic analysis in aged versus young FACS-sorted LSECs, with top hits indicated.The blue color indicates 44 negatively regulated genes, and the red color indicates 49 positively regulated genes.(B) The set of genes induced in aged LSECs was compared with a genome-wide ChIP-seq dataset using the Cscan program.Potential common transcriptional regulators of input genes are shown.(C,D) Shown are the relative mRNA expression levels of the ETS1 target genes Ptgis and Dusp6 in LSECs isolated from young and aged WT female mice.(E) Primary liver non-parenchymal cell (NPC) cultures were transfected with two independent siRNAs targeting ETS1.After 48 h the cells were treated with FeCit for 18 h.Shown are relative Bmp6 mRNA expression levels in FACS-sorted LSECs.(F-G) NPC cultures isolated from WT (female and male) and NRF2 KO (male) mice were treated with FeCit (1 mM) for 18 h.Ptgis and Dusp6 mRNA expression levels were determined in FACS-sorted LSECs.(H and K) Shown are Bmp6 or Aox1 mRNA levels in LSECs FACS-sorted from livers of young BALB/c female mice intravenously injected with FeCit (150 μg /mice) or citrate buffer.(I) Shown is a volcano plot of differentially regulated genes identified by the AmpliSeq transcriptomic analysis of FACS-sorted LSECs, isolated from iron citrate-and citrate-injected mice, with top hits indicated.The blue color indicates 47 negatively regulated genes, and the red color indicates 249 positively regulated genes.(J) The set of genes induced in iron citrate-exposed LSECs was compared with a genome-wide ChIP-seq dataset using the Cscan program.Potential common transcriptional regulators of input genes are shown.(L) Aox1 gene expression levels were measured in liver tissue isolated from young versus aged mice.Rpl19 served as a reference gene for gene expression analysis.Data are presented as mean ± SEM and each data point represents one biological replicate.Welch's unpaired t-test was used to determine statistical significance between the two groups, while two-way ANOVA with Tukey's Multiple Comparison test was used for panel G; *p < .05,**p < .01,and ***p < .001.[Color figure can be viewed at wileyonlinelibrary.com] enhancing oxidative stress and the Bmp6 response may depend on the activity of JNK or p38 MAPK.While we found that the hepatocyteimposed additional activation of Bmp6 is preserved in cells pretreated with the JNK kinase inhibitor (Figure S10H), we found that it is diminished in the presence of the p38 MAPK inhibitor compared to untreated cells (1.7-fold vs. 3.4-fold, respectively; Figure S10I).Consistently, we found that the two ETS1 target genes, Ptgis and Dusp6, were more robustly regulated in hepatocyte co-cultures compared to NPC cultures (Figure S10J,K) and this response was suppressed by the p38 MAPK inhibitor (Figure S10L,M).Finally, we also observed that the impact of hepatocytes on Bmp6 response is preserved in NRF2 KO mice (Figure S10N,O).In conclusion, our data imply that the co-culture of NPCs with hepatocytes modifies the response of LSECs to iron treatment by increasing their susceptibility to iron-imposed oxidative stress, likely by lowering the ferritin H levels, and as a result provoking more pronounced activation of p38 MAPK signaling.

| DISCUSSION
The anatomical location of LSECs between nutrient-rich blood and hepatocytes is optimal for the production of the angiokine BMP6 which senses iron levels to regulate hepcidin transcription. 31Both LSEC-intrinsic and hepatocyte-instructed extrinsic mechanisms were proposed to activate Bmp6 transcription in response to iron excess, but the exact nature of these regulatory pathways remains not completely understood.
Strikingly, our study demonstrated that NRF2 is dispensable for cell-autonomous iron-triggered Bmp6 induction in primary LSEC cultures, thus partially contradicting the conclusion of the previously reported study. 11However, Lim et al. presented data that irondextran-injected NRF2 KO mice still activate hepatic Bmp6 mRNA expression (especially in male mice), although to a lesser extent than WT animals.They further showed that this response was disrupted in NRF2 KO mice upon prolonged iron-rich feeding.Possibly the lack of LSEC responsiveness in the latter model may be explained by severe stress and necroinflammation imposed on the liver by the combination of massive iron loading and impaired antioxidant defense due to NRF2 deficiency, as indicated by disrupted sinusoidal architecture and injured hepatocytes, shown in this study. 11Finally, Lim et al.
employed studies in non-endothelial cells to demonstrate that RNAimediated silencing of NRF2 suppresses iron-driven Bmp6 induction.
We believe that the discrepancies between our findings and the model proposed by Lim et al. may be explained by the existence of alternative endothelial-specific regulatory mechanisms, such as ETS1, that can activate the Bmp6 promoter under iron-rich conditions.
The transcription factor ETS1 emerged as a candidate mediator of Bmp6 regulation through bioinformatic analysis of the LSEC transcriptomic data in the mouse model of aging and after intravenous iron citrate injection, an unbiased approach that was not used in previous studies.We believe that the activation of ETS1 in LSECs, possibly but not necessary, via stress inducible MAPKs, is a common denominator between the aging-triggered Bmp6 upregulation and its induction upon acute exposure to excessive iron.While literature reports ROS-associated p38 MAPK activation in the livers of approximately 2-year-old mice, 32 this kinase may be activated by a wide range of other extracellular factors. 33,34Furthermore, it should be emphasized that the ETS1 transcription factor is regulated by a plethora of posttranscriptional and posttranslational mechanisms, including activating and inhibitory phosphorylation, acetylation, SUMOylation, ubiquitination as well as protein-protein interactions. 35In sum, further research would be needed to investigate the exact mechanisms of LSEC ETS1 activation during aging.
ETS1 functions as a master endothelial transcriptional regulator.It is critical for vascular development during embryogenesis, 36 and has been shown to bind loci of VEGFA-responsive genes, amplifying their transcription during angiogenic stimulation. 37In 2023, upon preparation of this manuscript, Charlebois et al. reported data from single-cell RNA sequencing of liver cells derived from iron-depleted mice that were exposed to either Holo-transferrin injection or a high-iron diet. 38Using bioinformatic analysis of these data, the authors identified the transcriptional signatures indicative of MYC and NRF2 activation, particularly in response to excessive iron feeding.Interestingly, ETS1 and MYC were reported to overlap at endothelial promoters and mediate global gene expression changes via the release of paused RNA polymerase II. 37,39ture investigations would be needed to address whether MYC cooperates with ETS1 to regulate LSEC Bmp6 expression.(approx.threefold) induction of Bmp6, hallmarked by an ETS1-associated transcriptional signature.In sum, this implies that ETS1 may play a critical role in the early Bmp6 response to the iron stimulus, which may involve LSEC TFR1.Of note, our bioinformatic analysis identified another top candidate, MAFK, that may mediate global gene expression changes in LSECs upon short-term iron exposure.MAFK belongs to the family of small MAF proteins that can form heterodimers with other leucine zipper proteins, including c-Fos, a component of the AP-1 transcription factor. 40When this work was under review, another component of AP-1, c-Jun, was proposed as a Bmp6-regulating factor. 41Interestingly, while c-Jun is a major effector of the JNK kinase, 28 the p38 MAPK was shown to phosphorylate c-Fos. 42ETS1 and AP-1 were reported to synergistically drive transcriptional activation. 43Thus, it is plausible that MAFK, AP-1, and ETS1 may act in concert to trigger Bmp6 transcriptional induction, downstream of JNK and p38 MAPK activation.
Due to their well-established roles in mediating the cellular responses to oxidative stress, JNK and p38 MAPK are well-suited to mediate Bmp6 induction upon iron exposure. 27Studies involving chromatin conformation capture (3C) and co-immunoprecipitation would be needed to identify bona fide interaction between regions of the Bmp6 promoter which recruit phospho-ETS1 and phospho-p38 MAPK.It also remains an open question how JNK and p38 MAPK become activated upon iron-triggered oxidative stress.Two redox-sensitive factors acting upstream of these effector kinases are likely candidates.One is a thioredoxin-regulated signalosome composed of apoptosis signal-regulating kinase (ASK1), which activates p38 MAPK via phosphorylation of MKK3/6. 44Another mechanism might be mediated by MAP three kinase 1 (MTK1), a shared upstream activator of both JNK and p38 kinases via MKK3/4 and 6, which itself acts as a redox sensor. 45Noteworthy, the upstream ROS-dependent mechanisms that activate stress-inducible MAP kinases are likely distinct from those that operate in inflammation. 46is explains why Bmp6 mRNA expression levels are not induced by inflammatory cues. 47nally, our study adds valuable data to the still ongoing debate of whether Bmp6 induction in LSECs is cell autonomous 10 or requires the presence of hepatocyte-derived signals. 13Our data are partially consistent with both models, reporting a robust ROS-dependent induction of Bmp6 in NPCs cultures and an additional activation imposed on hepatocytes.Some discrepancies between our findings and the data of Colucci et al. 13

F I G U R E 1
Liver iron accumulation during aging increases liver sinusoidal endothelial cells Bmp6 expression independently of oxidative stress and NRF2-mediated signaling.(A) Non-heme liver iron levels and (B) transferrin saturation were measured in young (2-3 months) and aged (10-11 months) female C57BL/6J mice.(C) Total iron content in purified hepatocytes and liver non-parenchymal cells (NPCs) isolated from young and aged mice was assessed using a commercial assay kit.(D) Iron accumulation in liver tissue was visualized with DAB-enhanced Perls staining, bar = 100 μm.(E) Ferritin L (FTL) levels were determined in isolated hepatocytes by Western blot, while (F) FTL and (I) ferroportin (FPN) levels were measured in LSECs by flow cytometry.(H) The level of oxidative stress in LSECs was assessed by flow cytometry using a CellROX probe.(G, J-K) Shown are mRNA expression levels of Nqo1, Bmp6, and Hamp in the liver of young and aged mice.(L-P) Shown are mRNA expression levels of Bmp6, Hmox1, Nqo1, and Gclc in FACS-sorted LSEC populations of young and aged wild-type (WT) or NRF2 knockout (NRF2 KO) mice.Rpl19 was used as a reference gene for gene expression analysis.Data are presented as mean ± SEM and each data point represents one biological replicate.Welch's unpaired t-test was applied to determine statistical significance; ns-not significant, *p < .05,**p < .01,***p < .001,and ****p < .0001.LSEC, liver sinusoidal endothelial cells.[Color figure can be viewed at wileyonlinelibrary.com] next investigated the iron-triggered response of LSECs which were isolated in parallel from WT and NRF2 KO male mice.Strikingly, we found that induction of Bmp6, along with Hmox1 and Nqo1, the NRF2 target genes, 11 occurred independently of the NRF2 presence (Figure 2H-J), also reflecting the existence of other stress-associated regulatory mechanisms of these genes found in literature and the F I G U R E 2 Legend on next page.
Charlebois et al. further proposed that the transcriptional responses of LSECs to iron excess are primarily triggered by NTBI.This and other reports indicated that endothelial TFR1 has a relatively minor role in iron sensing. 10,38However, Charlebois et al. reported that both Holo-transferrin injection and a high-iron diet resulted in relatively mild TFR1-independent Bmp6 induction (up to twofold) at the 18-hour time point analyzed by this study.Strikingly, the authors showed but did not emphasize that ferric ammonium iron injection (for 5 h) led to the most robust induction of Bmp6 (approx.fourfold) and this response required endothelial TFR1.Likewise, a milder dose of ferric citrate injected into mice in our study triggered a clear F I G U R E 4 Induction of Bmp6 gene expression by iron involves c-Jun N-terminal kinases and engages ETS1-p38 mitogen-activated protein kinases axis activity.(A,B) Phosphorylation of JNK and p38 MAPK were determined by Western blot in liver non-parenchymal cell (NPC) cultures treated with FeCit (1 mM) for 2 and 4 h.(C-F) NPCs cultures were treated with JNK MAP kinase inhibitor SP600125 (20 μM) or p38 MAP kinase inhibitor SB202190 (10 μM) for 1 h before the FeCit (1 mM) administration for 18 h.Quantification of Western blot data was performed by densitometric analysis.(D) The level of oxidative stress in LSECs was assessed by flow cytometry using a CellROX probe and LSECs were FACS-sorted for gene expression analysis of (C) Bmp6, (E) Hmox1, and (F) Ptgis.Rpl19 was used as a reference gene for gene expression analysis.(G-I) Shown are chromatin immunoprecipitation (ChIP) results in NPC cultures treated with SB202190 (20 μM), Trolox (20 μM), or MitoTEMPO (500 μM) 1 h before the FeCit (1 mM) administration for 12 h.ChIP was performed with (G) anti-PolII antibody, (H) anti-phospho-ETS1 (pETS1) antibody and (I) anti-phospho-p38 MAPK (p-p38) antibody versus unspecific IgG binding.Proximal (Bmp6_prox) and distal (Bmp6_up) promoter regions of Bmp6 were amplified with qPCR.Data are presented as mean ± SEM and each data point represents a biological replicate.One-way ANOVA was used to determine statistical significance in panels A and B, while two-way ANOVA with Tukey's Multiple Comparison test was used for other panels; ns-not significant, *p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001.[Color figure can be viewed at wileyonlinelibrary.com] may arise due to differences in LSEC purification methods (we cultured LSECs with other NPCs and FACS-sorted them for analysis, whereas Colucci et al. applied two-step stringent isolation protocol before LSEC seeding) and thus the potential contribution of other NPCs (Kupffer or stellate cells) to Bmp6 transcriptional responses in our hands.Nevertheless, we propose that Bmp6 induction in the presence of hepatocytes partially requires p38 MAPK signaling.