Selective Hypoxia-Sensitive Oxomer Formation by FIH Prevents Binding of the NF-κB Inhibitor IκBβ to NF-κB Subunits

Abstract Pharmacologic inhibitors of cellular hydroxylase oxygen sensors are protective in multiple preclinical in vivo models of inflammation. However, the molecular mechanisms underlying this regulation are only partly understood, preventing clinical translation. We previously proposed a new mechanism for cellular oxygen sensing: oxygen-dependent, (likely) covalent protein oligomer (oxomer) formation. Here, we report that the oxygen sensor factor inhibiting HIF (FIH) forms an oxomer with the NF-κB inhibitor β (IκBβ). The formation of this protein complex required FIH enzymatic activity and was prevented by pharmacologic inhibitors. Oxomer formation was highly hypoxia-sensitive and very stable. No other member of the IκB protein family formed an oxomer with FIH, demonstrating that FIH-IκBβ oxomer formation was highly selective. In contrast to the known FIH-dependent oxomer formation with the deubiquitinase OTUB1, FIH-IκBβ oxomer formation did not occur via an IκBβ asparagine residue, but depended on the amino acid sequence VAERR contained within a loop between IκBβ ankyrin repeat domains 2 and 3. Oxomer formation prevented IκBβ from binding to its primary interaction partners p65 and c-Rel, subunits of NF-κB, the master regulator of the cellular transcriptional response to pro-inflammatory stimuli. We therefore propose that FIH-mediated oxomer formation with IκBβ contributes to the hypoxia-dependent regulation of inflammation.


Introduction
Decreased oxygen availability (hypoxia) and inflammation frequently co-occur with mutual functional effects. 1 Cells adapt to hypoxia via cellular oxygen sensors, including prolyl-4hydroxylase domain (PHD) proteins 1-3 and factor inhibiting HIF (FIH). 2 FIH was first described to catalyze the hydroxylation of an asparagine residue within a subunits of the heterodimeric hypoxia-inducible factor (HIF), [3][4][5][6] inhibiting the interaction of HIF with the transcriptional co-activators p300 and CBP histone acetyl transferases. 72][13] FIH has recently been linked to the regulation of inflammation in vivo.Treatment with the FIH-selective pharmacologic agent N-oxalyl-D-phenylalanine (NOFD) was protective against irradiation-induced injuries of the gastrointestinal tract and FIH deletion in the colon epithelium attenuated chronic colitis. 14,15We previously demonstrated that the enzymatic activities of FIH and PHD1 attenuate the major pro-inflammatory IL-1b signaling pathway. 12Nonetheless, the general relevance and the detailed FIH-dependent regulation of inflammation remains elusive.
FIH enzymatic activity is not limited to asparagine hydroxylation or HIF-a subunits. 16The first FIH substrates identified outside the HIF pathway were IjBa and p105, both members of the inhibitor of NF-jB (IjBs) protein family. 17The IjB protein family can be separated into three different groups: typical (IjBa, IjBb and IjBe), precursor (p105/IjBc and p100/ IjBd) and atypical (nuclear) IjBs (Bcl-3, IjBf, IjBNS, and IjBg). 18,19In the absence of a stimulus, NF-jB is retained in the cytoplasm by interaction with typical IjB proteins. 18ollowing an inflammatory stimulus, typical IjBs are phosphorylated and degraded, releasing NF-jB which then translocates into the nucleus and upregulates the transcription of specific genes. 20,21In addition to the hydroxylation of IjBa and p105, IjBe has previously been shown to interact with FIH. 22Of note, all members of the IjB protein family contain ankyrin repeat domains (ARDs) and proteins with ARDs represent the largest group among the known FIH substrates. 17,22,23Thus, there is strong link between FIH and IjB proteins, although the hydroxylation of IjB proteins is apparently not affecting IjBs' function. 17,24e recently reported that the deubiquitinase (DUB) ovarian tumor domain-containing ubiquitin aldehyde binding 1 (OTUB1) is a substrate for FIH-mediated hydroxylation of asparagine residue N22. 12,25In addition, FIH and OTUB1 form an oxygen-dependent denaturation-resistant (likely covalent) protein-protein complex catalyzed by FIH activity, which we refer to as "oxygen-dependent stable protein oligomer" (oxomer). 16,26,27FIH-OTUB1 oxomer formation is highly hypoxia sensitive and regulates OTUB1 enzymatic function. 26A mass spectrometry-based analysis identified 12 additional putative stable protein complexes formed by FIH, including ubiquitin-like modifier-activating enzyme 1 (UBA1) and IjBb. 26Interestingly, we had previously identified IjBb as interaction partner of FIH together with oxidation of aspartate residue 195, indicating a potential hydroxylation of IjBb by FIH. 12 Here, we investigated if FIH forms oxomer complexes with other proteins.Due to the existing functional links between FIH and inflammation, we focused on IjB proteins as putative substrates for oxomer formation.The identification of additional oxomer complexes and their characterization will increase our understanding of this novel type of oxygendependent signaling.Moreover, characterization of the functional interaction of FIH with proteins of pro-inflammatory signaling pathways will contribute to our understanding of the molecular mechanism(s) underlying the protective effect of pharmacologic FIH inhibition and genetic deletion in inflammatory diseases in vivo.

FIH forms an oxomer with IjBb but not with other IjB protein family members
Following the previous identification of 12 potential substrate proteins for oxomer formation with FIH, 26 we aimed to analyze these candidates in more detail and focused on IjBb and UBA1.IjBb was of special interest, as members of its protein family had been shown to be hydroxylated and/or to interact with FIH. 17,22,23Moreover, we had previously identified IjBb as interactor of FIH and we found a putative hydroxylation site within IjBb. 12 UBA1 was selected for further investigation because it plays a pivotal role in protein ubiquitination as E1 enzyme, initiating the enzymatic ubiquitination cascade. 28,29We thus assessed the oxomer formation between ectopically expressed FIH and IjBb (Figure 1A) and between FIH and UBA1 (Figure S2) by immunoblotting.Oxomer formation between FIH and OTUB1 served as positive control (Figure 1A). 26Ectopic co-expression of UBA1 or IjBb with FIH led to the formation of denaturing conditionresistant (SDS-PAGE) complexes at molecular weights that were consistent with the covalent addition of UBA1 or IjBb, respectively, to FIH (Figure 1A and Figure S2).Moreover, a stable FIH-IjBb complex was also formed in HepG2, Hep3B and MCF7 cells (Figure 1B), suggesting a general occurrence of the observed oxomer formation.Interestingly, the ratio between the FIH-IjBb oxomer and IjBb monomer differed between the cell lines with the highest relative FIH-IjBb oxomer levels being observed in MCF-7 cells (Figure 1B).
Next, we assessed if FIH-IjBb oxomer formation was also observable between endogenous proteins and whether it was dependent on FIH enzymatic function.Following IP of endogenous IjBb, a signal was detected at a molecular weight that corresponded to the predicted FIH-IjBb oxomer (Figure 1C and Figure S3A).This signal was abolished following FIH inhibition by DMOG treatment or FIH gene deletion (Figure 1C and Figure S3A).Thus, also endogenous FIH and IjBb form a denaturing condition-resistant oxomer which depends on FIH enzymatic activity.
The successful validation of FIH-IjBb oxomer formation raised the question whether other IjB family members also served as FIH substrates for oxomer formation.To investigate this possibility, FIH was ectopically co-expressed with either IjBa, IjBb, Bcl-3, IjBd, IjBe, p105 or p100 and it was assessed if any higher molecular weight signals could be detected that were sensitive to pharmacologic FIH inhibition (Figure 1D) or FIH deletion (Figure S3B).Although several IjB family members had previously been shown to serve as hydroxylation substrates and/or interaction partners of FIH, 17,22,23 none of the other IjB proteins formed an oxomer with FIH (Figure 1D and Figure S3B), demonstrating that FIH-mediated oxomer formation is highly selective for IjBb.
Because FIH-IjBb oxomer formation required FIH activity, the sensitivity of FIH-IjBb oxomer formation to different oxygen availability (18.5-1% O 2 ) was investigated.Interestingly, FIH-IjBb oxomer formation was highly sensitive to hypoxia, with an EC 50 of 7.6% O 2 compared to an EC 50 for HIF-1a stabilization of 2.0% O 2 within the same cells (Figure 2D and E).

The FIH-IjBb oxomer is an extraordinarily stable protein complex
Given the resistance of the FIH-IjBb oxomer against the denaturing conditions during immunoblotting, we next characterized the stability of this oxomer in more detail.Recently, a new type of covalent bond between a lysine and a cysteine residue has been described, called a NOS bridge. 33The NOS bridge can be disrupted under very strong reducing conditions (1-5 mM dithiothreitol, DTT).To investigate if FIH and IjBb are linked through a NOS bridge within the observed oxomer, samples were treated with different reducing agents at even higher concentrations than those used for disrupting the NOS bridge (Figure 3A).None of the used concentrations of DTT or b-mercaptoethanol disrupted the FIH-IjBb oxomer, indicating that the linkage between FIH and IjBb is not a NOS bridge or any other type of bond that is susceptible to the reducing conditions described.
Next, we analyzed the degradation and stability of the FIH-IjBb oxomer.To assess if the oxomer was degraded by the proteasome, cells were allowed to form the oxomer for 24 h and subsequently treated with the proteasome inhibitor MG132 for 21 h (Figure 3B).Proteasome inhibition strongly increased FIH, IjBb and the oxomer levels (Figure 3B), also demonstrating that the oxomer is degraded via the proteasome.In a subsequent experiment, the stability of the oxomer was analyzed.Following oxomer formation for 22 h, cells were exposed to normoxia or hypoxia (0.2% O 2 ) for the indicated time periods to prevent any additional oxomers to be formed (Figure 3C).In normoxia, no change of the oxomer protein level was observed over the entire 48 h (Figure 3C  and D).In hypoxia, a first significant reduction of the oxomer was detected only 24 h after the start of the hypoxia exposure and reached approximately 50% of the initial levels at around 48 h following the onset of hypoxia (Figure 3C and D).

The IjBb loop between ankyrin repeat domains 2 and 3 is necessary for oxomer formation
We next aimed to identify the IjBb amino acid targeted by FIH for oxomer formation.FIH is known to hydroxylate asparagine residues and the FIH-OTUB1 oxomer is formed through asparagine 22 (N22) of OTUB1. 4,26IjBa has so far been the best characterized FIH target protein outside the HIF pathway, 17,34 it is hydroxylated by FIH on the two asparagine residues Asn-244 and Asn-210 and it is closely related to IjBb. 17 We therefore aimed to compare the IjBa and IjBb structures to obtain insights into the potential part of IjBb that might be targeted by FIH enzymatic activity.However, to the best of our knowledge, the protein structure of human IjBb has not been experimentally solved to date.6][37] Figure 4A highlights the asparagine residues of IjBa that are hydroxylated by FIH and which are located within a loop between ARDs 2 and 3. 17 Interestingly, one IjBb asparagine residue (N266) was positioned in a comparable area of IjBb as the hydroxylated asparagine residues in IjBa (Figure 4A).Although this indicates that N266 may be the targeted amino acid residue within IjBb, we decided to point mutate all of the five IjBb asparagine residues to alanine.In addition, one aspartate residue was also point mutated to alanine (D195A), as previous mass spectrometry analyses had shown its oxidation and because FIH also hydroxylates aspartate residues. 12,38AlphaFold-mediated structure prediction indicated that the introduced point mutations do not alter the structure of IjBb (Figure S4A).Unexpectedly, none of the introduced IjBb point mutations prevented FIH-IjBb oxomer formation (Figure 4B).The higher molecular weight signal of the oxomer doublet showed some variability within its intensity (Figure 4B), but this was likely due to differences in the ectopic expression efficiency of the mutated IjBb (Figure 4B).Thus, in contrast to the asparagine residue in the FIH-OTUB1 oxomer, FIH appears to utilize a different aa residue for oxomer formation with IjBb.
Next, we analyzed whether IjBb is hydroxylated by FIH via mass spectrometry (MS), because for OTUB1 the same asparagine residue that is involved in oxomer formation is also hydroxylated in monomeric OTUB1. 26We identified IjBß peptides containing four of the five asparagine residues.Despite detecting numerous oxidations on methionines, none of the asparagines were deemed to be hydroxylated (Supplemental Table S1).The one unidentified asparagine resides in the Nterminal region and was not accessible by standard MS workflows because the tryptic digest generated a peptide that was too large for MS identification.Nevertheless, it is highly likely that IjBß is not hydroxylated as none of the sequences surrounding any of the IjBß asparagines matches the wellcharacterized FIH consensus motif Lx(6)[VI]N. 39Therefore, we systematically screened for the IjBb moiety that was involved in FIH-IjBb oxomer formation.First, we assessed the N-(ARDs 1-3, aa residues 1-166) or C-terminal (ARDs 4-6, aa residues 167-356) parts of IjBb (Figure 4C).According to AlphaFold predictions, these truncations largely maintained the corresponding protein structures (Figure S4B).While IjBb ARDs 1-3 formed an oxomer with FIH with the predicted molecular weight, there was no oxomer formed with ARDs 4-6 (Figure 4C).
To identify the IjBb residues involved in oxomer formation, vectors were constructed containing IjBb ARD 1-3 with truncations of 10 aa increments from the C-terminal end (Figure 4D).Truncated IjBb fragments containing the Nterminal 126 aa or less were no longer able to form an oxomer with FIH (Figure 4D).Next, IjBbs with consecutive deletions were generated in otherwise full-length IjBb with overlapping steps of 5 aa, surrounding the identified area (Figure 4E).While deletion of IjBb aa 127-136, 132-141 and 137-146 did not interfere with oxomer formation, the absence of IjBb aa 117-126 and 122-131 abrogated oxomer formation (Figure 4E).Therefore, the IjBb aa 122-126 (VAERR) region, common to both deletion constructs, is necessary for oxomer formation between FIH and IjBb.These residues are located in the loop between the IjBb ARDs 2 and 3 (Figure S4C and D).Of note, the IjBb peptide containing VAERR was detected in the MS analyses, but no relevant amino acid residue was deemed to be hydroxylated within this peptide.Overall, the IjBb VAERR sequence is necessary for FIH-IjBb oxomer formation, but it remains unclear if any of the included amino acid residues is directly targeted by FIH enzymatic activity.

FIH-IjBb oxomer formation prevents IjBb from binding the major NF-jB subunits p65 and c-Rel
It has previously been suggested that IjBa can scavenge FIH from HIF-1a, reducing or preventing FIH-mediated hydroxylation of HIF-1a, and thus increasing HIF-1 transcriptional activity. 17Therefore, we aimed to investigate a potential effect of IjBb on FIH activity through oxomer formation by employing a HIF-dependent firefly luciferase reporter gene assay. 40,41EK293 cells were transfected with various IjB proteins for 24 h and treated with FG-4592 for an additional 24 h to specifically inhibit the PHDs to increase HIF activity without pharmacologically affecting FIH. 27FG-4592 treatment increased HIF-dependent firefly luciferase activity (Figure 5A).Ectopic expression of IjBa and IjBe significantly increased HIF activity likely by reducing FIH activity towards HIF (Figure 5A), as previously described. 17,22Interestingly, p105 had no additional effect compared to FG-4592 treatment alone (Figure 5A), although p105 is hydroxylated by FIH. 17 No other IjB protein family member, including IjBb, affected HIF activity.These results indicate that the effect of IjBa and IjBe   17 are indicated in dark blue, all asparagine residues of IjBb are highlighted in red.An aspartic acid residue that had previously been found to be oxidized is highlighted in magenta. 12 towards HIF activity is selective and not a general effect of all FIH target proteins or all ARD-containing FIH substrates.
The main function of IjBb is to bind to NF-jB transcription factor dimers, which sequesters them in the cytosol and regulates NF-jB-mediated gene expression. 20,21Thus, we investigated whether IjBb retains its capability to bind NF-jB proteins within the oxomer.Specifically, the interaction between IjBb and p65 as well as c-Rel was assessed, as these NF-jB subunits are most commonly bound by IjBb. 42IjBb-FLAG was ectopically expressed in HEK293 cells alone or in combination with FIH-V5, and immunoprecipitated using anti-V5 or anti-FLAG antibodies, respectively.The FLAG IP pulled down the FIH-IjBb oxomer as well as monomeric IjBb, whereas the V5 IP pulled down the oxomer as well as FIH but no monomeric IjBb (Figure 5B), indicating that FIH was bound as homodimer to IjBb as observed within the FIH-OTUB1 oxomer. 26p65 and c-Rel were both co-immunoprecipitated together with IjBb-FLAG, but not in the FIH-V5 IP (Figure 5B and 5C), which contained IjBb only as oxomer.These results suggest that oxomer formation of IjBb with FIH prevents IjBb from binding p65 and c-Rel and therefore removes IjBb from the functionally relevant pool of IjB proteins for the regulation of NF-jB.

Discussion
4][45][46] Tissue hypoxia is commonly observed in inflamed tissue areas and an inflammatory response can occur in hypoxic tissues. 9,46][10] Our understanding of the molecular mechanisms underlying this mutual interplay, however, is in its infancy.FIH may be part of the regulatory network connecting hypoxia and inflammation, as FIH alters IL-1b-induced NF-jB activity and is relevant for the inflammatory response in the gut in vivo. 12,14,15onetheless, the extent of contribution of FIH activity to inflammation as well as the molecular mechanisms that connect FIH and inflammation are unclear.
Previously, we reported that FIH forms an oxomer with OTUB1. 26Here, we demonstrate that FIH forms a similar oxomer specifically with IjBb.FIH enzymatic activity was necessary for the FIH-IjBb complex formation, which was prevented by pan-hydroxylase inhibitors and FIH mutations.The oxomer was formed in several different cell lines and by endogenous proteins.The endogenous oxomer was observed at relatively low levels compared to the IjBb monomer.However, this assessment was performed in HEK293 cells and may differ in other cell lines.Interestingly, we observed (with ectopically expressed proteins) that the oxomer level relative to monomeric IjBb is higher in MCF7 than in HEK293, Hep3B and HepG2 cells, indicating that there might be cell types in which oxomer formation is favored.Moreover, following a pro-inflammatory stimulus, IjBb is degraded and re-expressed. 47,48We have previously reported that oxomer formation occurs co-translationally. 26Thus, following a proinflammatory stimulus, (endogenous) oxomer formation may be increased due to the strong induction of IjBb expression and translation.This will be analyzed in the future.
Interestingly, IjBb was the only protein of the IjB protein family that was targeted by FIH for oxomer formation, although all members contain ARDs which are prominent FIH substrates. 17,22,23Within our initial screen for FIH target proteins for oxomer formation, IjBb was the only identified ARD-containing protein. 26Together, these results demonstrate that FIH-mediated oxomer formation is highly selective and indicate that ARDs are not a defining characteristic of substrate proteins for FIH-dependent oxomer formation.
The FIH-mediated FIH-IjBb oxomer formation demonstrated a higher hypoxia sensitivity than the PHD-mediated stabilization of HIF-1a.Moreover, this sensitivity was even higher than the sensitivity of FIH-mediated HIF-1a hydroxylation, which is known to be less susceptible to hypoxia than PHD-mediated hydroxylation. 49These results are remarkable and comparable to our previous findings regarding the hypoxia sensitivity of the FIH-OTUB1 oxomer formation. 26This further strengthens the hypothesis that oxomer formation serves as an alternative cellular mechanism to fine-tune the adaptation to hypoxia.Moreover, it indicates that the hypoxia sensitivity of oxomer formation is largely defined by yet unknown characteristics of FIH and less by its protein substrates.In addition, the FIH-IjBb oxomer was very stable and only started to noticeably decline 24 h after the onset of hypoxia.These results are consistent with the FIH-OTUB1 oxomer and support the conclusion that FIH-dependent oxomers may serve for signaling long-term rather than short-term fluctuations in local oxygen availability. 26s previously observed for the FIH-OTUB1 oxomer, 27 the FIH-selective inhibitor DM-NOFD failed to prevent FIH-IjBb oxomer formation.Of note, both oxomers showed a higher hypoxia sensitivity than FIH-mediated HIF-a asparagine hydroxylation, 26 indicating that the sensitivity of FIH for cosubstrate binding depends on the protein substrate.This may also extend to co-substrate mimetics, such as DM-NOFD.To test this hypothesis, it would be necessary to assess the DM-NOFD sensitivity of FIH-dependent HIF-a asparagine hydroxylation and oxomer formation within the same lysates.However, a reliable antibody detecting HIF-a asparagine hydroxylation is currently not commercially available.
The connection between FIH and IjBb in the FIH-IjBb oxomer resisted high concentrations of reducing agents and boiling, suggesting a likely covalent bond, and indicating that the oxomer is formed independently of cysteine residues.FIH hydroxylates asparagine residues 4 and the FIH-OTUB1 oxomer is formed through asparagine 22 (N22) of OTUB1.However, FIH is known to be promiscuous 16 and hydroxylates aspartate, 38 histidine, 50 and tryptophan residues as well. 51Interestingly, we observed that point mutations of each asparagine within IjBb did not interfere with oxomer formation, demonstrating that the FIH-IjBb oxomer was not formed via an IjBb asparagine residue.IjBb-targeted MS analysis did not identify any potential hydroxylation of IjBb, indicating that the previously identified oxidation of D195 likely occurred by random reaction with oxygen during sample preparation and not by enzymatic activity. 12In addition, IP of FIH pulled down the FIH-IjBb oxomer, but not the IjBb monomer.Together, these results strongly indicate that IjBb is not hydroxylated by FIH.Previously, in an in vitro decarboxylation assay using purified FIH and IjBb, no FIH-dependent activity toward IjBb was detected, 17 further supporting that IjBb is not a hydroxylation substrate of FIH.Nonetheless, the results obtained by the in vitro assay are not contradicting our observation of the FIH-IjBb oxomer formation, because such oxomers are likely cotranslationally formed. 26Assuming that also the FIH-IjBb oxomer is formed cotranslationally, such a formation would not be detected by an assay that utilizes fully folded proteins.
Analyzing possible functional effects of the FIH-IjBb oxomer, the FIH-dependent regulation of HIF transactivation activity appeared to be unaffected in a HIF-dependent reporter gene assay.In the same experiment, ectopic expression of IjBa and IjBe increased HIF-dependent firefly luciferase activity, which is in agreement with a previous report that IjBa can inhibit FIH activity toward HIF-1a. 52To our knowledge, this is the first report that also IjBe -but no other IjB protein family member -affects HIF activity.
The main function of IjBb is to bind to NF-jB transcription factor dimers (mostly containing c-Rel and p65) and to retain them in the cytosol, preventing gene expression enhancement by these specific NF-jB dimers. 42It is therefore of interest that IjBb complexed with FIH as oxomer did not retain its ability to interact with the NF-jB subunits p65 and c-Rel.Consequently, oxomer formation may increase NF-jBdependent transcription by preventing p65 and c-Rel binding to IjBb.Such a regulation would be specific to NF-jB dimers involving p65 and/or c-Rel and would thus not necessarily affect the overall NF-jB-dependent gene expression, as 13 different NF-jB dimers have (so far) been demonstrated to exist. 53Further investigations will be necessary to determine the extent of the contribution of FIH-IjBb oxomer formation to the regulation of pro-inflammatory gene expression via inhibition of IjBb binding to NF-jB dimers.
Recently, it has been reported that FIH expression is increased in inflamed renal tissue during chronic kidney disease (CKD). 54It would therefore be of interest to test for the existence and potential regulation of the FIH-IjBb oxomer in healthy and diseased renal tissues, as an enhanced FIH-IjBb oxomer formation may contribute to increased inflammation.However, it is currently unclear to which extent oxomers exist in healthy or diseased tissues.
In summary, FIH forms a previously unknown oxomer with IjBb in a likely covalent manner and with high hypoxia sensitivity.FIH-mediated sequestration of IjBb within an oxomer does not affect the total FIH enzymatic activity present within a cell, but it prevents the complexed IjBb from inhibiting specific NF-jB dimers.Therefore, the FIH-IjBb oxomer formation and/or its hypoxic inhibition may contribute to the interplay between hypoxia/FIH and inflammation.

Cell culture and transient transfection
Human HEK293 (embryonic kidney), HepG2 and Hep3B (hepatocellular carcinoma), and MCF7 (breast adenocarcinoma) cell lines were cultivated in high glucose Dulbecco's modified Eagle's medium (Sigma-Aldrich, USA) with supplementation of 100 U/mL penicillin and 100 lg/mL streptomycin (Sigma-Aldrich) as well as 10% heat-inactivated fetal calf serum (Gibco, USA).Transient transfection of plasmids was performed with lipofectamine 2000 reagent according to the manufacturers' protocol (Invitrogen, USA).

RNA isolation and RT-qPCR
RNA was extracted by the guanidinium thiocyanate-acid phenol-chloroform method as described earlier. 55omplementary DNA (cDNA) was obtained by reverse transcription (RT) with AffinityScript transcriptase (Agilent, Santa Clara, CA, USA) and quantitated using the SYBR Green qPCR reagent kit (Kapa Biosystems, London, UK) and a MX3000P light cycler (Agilent) using standard curves.Transcript levels were normalized to the levels of human ribosomal protein L28 mRNA.Primer sequences are listed in Supplemental Table S2.
Incubations in hypoxia were performed using the InvivO 2 400 humidified cell culture workstation (Baker Ruskinn, Bridgend, South Wales, UK) operated with O 2 or 1% O 2 , 5% CO 2 at 37 � C as previously described, or in humidified oxygen-regulated cell culture incubators (Binder, Tuttlingen, Germany) operated with 2-8% O 2 and 5% CO 2 at 37 � C. 57 "normoxia" refers to the air oxygen level in the gas phase within a humidified cell culture incubator at 500 m altitude (18.5% O 2 ). 58

Reporter gene assays
HEK293 cells were cotransfected with pH3SVL, encoding the firefly (Photinus pyralis) luciferase reporter gene under the control of a SV40 promoter and concatamerized transferrinderived hypoxia response elements (HREs) (total of six HIF binding sites) and pRL-SV40, encoding a constitutively expressed Renilla (Renilla reniformis) luciferase reporter construct driven by the SV40 promoter, in combination with a member of the IjB protein family (IjBa, IjBb, Bcl-3, IjBd, IjBe, p105, p100) or empty vector. 40,41Twenty-four hours after transfection, cells were treated with 0.1 mM prolyl hydroxylase inhibitor FG-4592 (roxadustat) or vehicle (DMSO) for 24 h.Dual-Luciferase Reporter Assay was used to determine firefly and Renilla luciferase bioluminescence according to the manufacturers' instruction (Promega, Madison, Wisconsin, USA).In brief, cells were washed with PBS and lysed with Passive Lysis Buffer (Promega) for 10 min at room temperature.Lysates were mixed with equal volumes of Luciferase Assay Reagent II and luminescence was measured by a microplate luminometer (Berthold Technologies, Bad Wildbach, Germany).Subsequently, freshly combined Stop & Glo reagent (Promega) was added and the activity of Renilla luciferase was determined.

Protein structure prediction
Predictions of protein structures were created with ColabFold using the default settings, which resulted in five structure predictions, ranked by Predicted Aligned Error (PAE). 59Rank 1 model was used for further analysis and representation (https://colab.research.google.com/github/sokrypton/ColabFold/blob/main/AlphaFold2.ipynb).

Statistical analysis
For the analysis of the significance of difference between two data points, Student's t test was applied.For comparison of more than two data points, one-way or two-way ANOVA followed by Tukey's post-test was applied.P values < 0.05 were considered statistically significant.The sample sizes for each experiment have been indicated in the corresponding figure legends.

Figure 5 .
Figure 5. FIH-IjBb oxomer formation prevents IjBb from binding NF-jB subunits.(A) HEK293 cells were transiently cotransfected with a HIF-driven luciferase plasmid (pH3SVL) together with a constitutive Renilla luciferase plasmid (pRL-SV40) as well as plasmids expressing the indicated IjB protein family member 24 h prior to treatment.FG-4592 treatment was applied for 24 h in a concentration of 0.1 mM.Efficient transfection and FG-4592 treatment were determined by immunoblotting (Figure S5).Significance was assessed by two-way ANOVA with Tukey's correction; � P < 0.05, �� P < 0.01.(B) Immunoprecipitation (IP) analysis of the interaction of IjBb and the FIH-IjBb oxomer with the NF-jB subunits c-Rel and p65.(C) Quantification of the results shown in B. The p65 and c-Rel levels were normalized to the levels of total bait/FLAG pulldown (IjBb or IjBb þ oxomer), respectively.Statistical analysis was performed by Student's t test.� P < 0.05; F, FIH-V5; I, IjBb-FLAG.Data are shown as mean ± SD.Data are representative of three independent experiments.