Structure and Function of RhoBTB1 Required for Substrate Specificity and Cullin-3 Ubiquitination

Abstract We identified Rho-related BTB domain containing 1 (RhoBTB1) as a key regulator of phosphodiesterase 5 (PDE5) activity, and through PDE5, a regulator of vascular tone. We identified the binding interface for PDE5 on RhoBTB1 by truncating full-length RhoBTB1 into its component domains. Co-immunoprecipitation analyses revealed that the C-terminal half of RhoBTB1 containing its two BTB domains and the C-terminal domain (B1B2C) is the minimal region required for PDE5 recruitment and subsequent proteasomal degradation via Cullin-3 (CUL3). The C-terminal domain was essential in recruiting PDE5 as constructs lacking this region could not participate in PDE5 binding or proteasomal degradation. We also identified Pro353 and Ser363 as key amino acid residues in the B1B2C region involved in CUL3 binding to RhoBTB1. Mutation of either of these residues exhibited impaired CUL3 binding and PDE5 degradation, although the binding to PDE5 was preserved. Finally, we employed ascorbate peroxidase 2 (APEX2) proximity labeling using a B1B2C–APEX2 fusion protein as bait to capture unknown RhoBTB1 binding partners. Among several B1B2C-binding proteins identified and validated, we focused on SET domain containing 2 (SETD2). SETD2 and RhoBTB1 directly interacted, and the level of SETD2 increased in response to pharmacological inhibition of the proteasome or Cullin complex, CUL3 deletion, and RhoBTB1-inhibition with siRNA. This suggests that SETD2 is regulated by the RhoBTB1–CUL3 axis. Future studies will determine whether SETD2 plays a role in cardiovascular function.


Introduction
Hypertension (HTN) or high blood pressure (BP) is among the most pr ev alent factors underl ying car dio vascular impairments.
As r e ported by the American Heart Association, HTN is responsible for 69% of first heart attack, 77% of first stroke, and 74% of chronic heart failure. 1 Data from the Centers for Disease Control and Pr ev ention suggest that nearl y half of the US population has HTN and only 24% of hypertensive subjects have their BP under control. 2 The lifetime risk of heart failure is also twice in hypertensi v e subjects as compared to normotensive subjects. 3 In 2019, HTN w as the primar y or contributing cause of mortality of more than 0.5 million US adults. As av era ged ov er 12 yr fr om 2003 to 2014, the cost of HTN treatment is about $131 billion annually, which is expected to reach $1.1 trillion by 2035. 4 , 5 Consequently, it is essential that continued resear c h identifies new pathways and potential therapeutic targets to manage HTN.
Per oxisome pr oliferator-acti v ated r ece ptor γ (PPAR γ ) is a ligand-acti v ated transcription factor, which controls both the bioav aila bility of nitric oxide and the r esponsi v eness of the vascular smooth muscle to endothelial-deri v ed nitric oxide, and thus plays an important regulator of blood vessel tone. 6 Previous genetic studies have established the importance of PPAR γ in car dio vascular diseases, including HTN . [7][8][9] Early studies suggested that PPAR γ may mediate cardiopr otecti v e effects on the v asculatur e. 10 Mor e r ecentl y, expr ession of dominant negati v e PPAR γ isoforms in either endothelium or vascular smooth muscle caused an enhanced susceptibility to, or directly caused, vascular dysfunction and HTN. [11][12][13] But, the generalized acti v ation of PPAR γ by high affinity pharmacological agonists is associated with adverse car dio vascular outcomes. 14 The intent of our resear c h is to optimize the adv anta geous outcomes of PPAR γ acti v ation while reducing its unfavorable impacts by focusing on transcriptional targets of PPAR γ .
We discov er ed that Rho-r elated BTB domain 1 (RhoBTB1) is an important PPAR γ target gene in vascular smooth muscle cells. 12 Induction of RhoBTB1 expression in mice carrying a smooth muscle specific dominant negative mutation in PPAR γ r ev ersed HTN, v ascular dysfunction, and arterial stiffness. 15 Inducib le expr ession of RhoBTB1 in a model of pre-existing angiotensin-II-dependent hypertension quickly reversed arterial stiffness. 16 RhoBTB1 is a broad complex, tr amtr ac k, and bric a brac (BTB)-domain containing protein, which acts as a substrate adaptor for the Cullin-3 (CUL3) ubiquitin ligase, which plays a role in proteasomal de gr adation of target proteins. 17 , 18 CUL3 serves as a scaffold where components of the Cullin Ring Ligase 3 (CRL3) complex can assemble. We pr eviousl y determined that phosphodiesterase 5 (PDE5) is an important target of RhoBTB1 in vascular smooth muscle cells. 15 PDE5 plays a crucial role in the regulation of cyclic 3 ,5 -monophosphate (cGMP), and consequently, the activity of nitric oxide in v ascular m uscle is tightl y r egulated by RhoBTB1-CUL3-PDE5 signaling.
RhoBTB1 is a multi-domain protein comprised of an Nterminal GTPase domain, a "sandwiched" proline-rich region, two BTB domains, and a C-terminal (CT) domain. 18 The BTB domains are known to function as mediators of pr otein-pr otein interactions and serve as the distinctive feature of adaptor proteins that direct protein substrates to war ds ubiquitination via the CRL3 complex. 19 RhoBTB1 interacts with CUL3 through its first BTB domain, and pr esuma b l y to substrates through its CT domain. 20 However, the mechanisms causing recognition of specific substrates, and the range of substrates for this and many other BTB-domain containing proteins are poorly understood. Here , w e gener ated truncations of RhoBTB1 and assessed the domain r equir ements to bind to PDE5. We also identified r esidues r equir ed for the inter action betw een RhoBTB1 and CUL3 and showed they are required for CUL3-mediated de gr adation of PDE5, but not the binding of RhoBTB1 to PDE5. Finally, we tested the hypothesis that the same RhoBTB1 domains r equir ed to bind PDE5 could be used as bait to capture other RhoBTB1-binding proteins, some of which might be CUL3 targets. We validated several RhoBTB1 targets and showed that SET domain-containing 2 (SETD2) is regulated by proteasomal de gr adation via a CUL3 and RhoBTB1-dependent mechanism.

Transfection, Site-Directed Mutagenesis, and RNA Interference
Deletion mutants encoding Myc-tagged RhoBTB1 domains and His-ta gged PDE5 wer e cloned in pcDNA3.1 mammalian expr ession vector (V79020, Invitrogen, Thermo Fisher Scientific, USA) using BamH1 and EcoRV restriction sites. CUL3 was cloned in the pcDNA3.1 mammalian expression vector using Kpn1 and EcoRV restriction sites. For proximity labeling, APEX2-NES w as excised fr om the pcDNA3-APEX2-NES v ector (49386, Addgene, USA) using Not1 and Xho1 restriction sites and recloned downstream of B1B2C construct using the glycineserine linker GGSSGGSS (encoded by GGA GGCTCCTCA GGA G-GTTCGTCT) in the pcDNA3.1 vector. Site-directed mutagenesis tec hniques w ere used to gener ate all point mutants in the B1B2C domain using the QuickChange II Site-Directed Mutagenesis Kit (200523, Agilent, USA) according to man ufactur er's instructions. Sequences of the primers used in site-dir ected m uta genesis, real time RT-PCR, and siRNA for RhoBTB1 are listed in Supplementar y Ta b les S1-S3. All plasmid sequences wer e v erified using DNA sequencing (Retrogen Inc., USA). For expression of Myc-B1B2C and His-PDE5, HEK293 cells were transfected using Lipofectamine LTX-PLUS r ea gent (A12621, Invitrogen, Thermo Fisher Scientific, USA) according to the manufactur er's pr otocol . All siRN A duplexes w ere tr ansfected using Lipofectamine RNAiMax transfection r ea gent (13778075, Invitrogen, Thermo Fisher Scientific, USA) according to the man ufactur er's protocol.

Ubiquitination and Cycloheximide Pulse-Chase Assay
For B1B2C mediated in vi v o ubiquitination of PDE5, transfected cells wer e tr eated with MLN4924 or MG132 immediatel y after transfection with appropriate constructs, and the cell lysates wer e pr e par ed in RIPA l ysis buffer to sustain denaturing environment. The pull-down was performed, immunoprecipitates wer e r esolv ed, transferr ed to nitr ocellulose membranes, and the b lots wer e dev eloped using the methodology mentioned a bov e. For the cycloheximide pulse-chase assay, the cells were co-transfected with His-PDE5 and increasing concentrations of Myc-B1B2C for 16 h and then cells were treated with 100 μm cycloheximide (C7698, Sigma-Aldrich, USA) for another 8 h. Cell l ysates wer e pr e par ed in RIPA buffer supplemented with pr otease inhibitors and 1 m m PMSF (36978, Thermo Fisher Scientific, USA). Proteins were resolved by SDS-polyacrylamide gel electr ophor esis and transferr ed to a nitr ocellulose membrane as described a bov e. Pr otein lev els wer e normalized with gl yceraldeh yde 3-phosphate deh ydrogenase (GAPDH), and western blot images were analyzed using ImageJ and densitometric data were plotted using Prism 9.4.1 softw ar e (GraphPad, USA).

Real-time RT-PCR
HEK293 cells were transfected with appropriate constructs for 16 h. Cells were then treated with 5 μm of Actinomycin D (SBR00013, Sigma-Aldrich, USA) for another 8 h and then suspended in Trizol r ea gent (15596018, Invitr ogen, Thermo Fisher Scientific, USA). Total RNA was isolated, and 1 mg of RNA w as r ev ersed transcribed using 1 U of Superscript III (18080044, Invitrogen, Thermo Fisher Scientific, USA) in 20 μL reactions. Reactions were incubated at 50 • C for 5 min, 4 • C for 3 min, 55 • C for 45 min, and finally at 70 • C for 15 min for inacti v ation. The cDNA was diluted 1:40, and gene expression was measured using Fast SYBR Green Gene Master Mix (4385612, Applied Biosystems, Thermo Fisher Scientific, USA). The relati v e lev els of mRNA expr ession wer e normalized to GAPDH. Data wer e anal yzed using the 2 − CT method to calculate foldchanges r elati v e to the cells transfected with empty v ector. The sequence of primers used are listed in Supplementary Ta b le S2.

Immunofluorescence
HEK293 cells grown in 12-well chamber slides (81201, Ibidi, Germany) were either transfected with Myc-B1B2C and His-PDE5 or co-transfected with both Myc-B1B2C and His-PDE5. At 24 h post transfection, cells were fixed with 4% paraformaldehyde (PFA; J61899.AK, Thermo Fisher Scientific, USA) and permeabilized with 0.1% Triton X-100 in blocking solution (0.5% BSA in PBS) follow ed by bloc king for 1 h with 0.5% BSA in PBS, and then incubated overnight at 4 • C with mouse monoclonal anti-His (1:100) and rabbit monoclonal anti-Myc (1:100) primary antibodies followed by 3 washes with PBS. Cells were then incubated with Alexa Fluor goat anti-mouse 488 and Alexa Fluor 568 goat anti-ra bbit secondar y antibodies for 1 h at r oom temperatur e followed by 3 washes with PBS. Glass slides were mounted using ProLong Gold Antifade Reagent (P10144, Invitrogen, Thermo Fisher Scientific, USA) containing DAPI. Images were acquired with a 40X objecti v e using a confocal laser scanning microscope (LSM510; Zeiss, Oberkochen, Germany) and analyzed using the Aim 4.2 softw ar e LSM510.

Pr o ximity Ligation Assay (PLA)
HEK293 cells were seeded at a seeding density of 12 500 cells/well in 12-w ell c hamber slides and grown for 2 days. After 2 days, cells w ere tr ansfected with Myc-B1B2C and His-PDE5 or cotransfected with both Myc-B1B2C and His-PDE5 clones for 24 h following a published PLA protocol. 22 The cells were washed 3 times with PBS, fixed in 4% PFA in PBS for 10 min, permeabilized by ice-cold methanol for 20 min, washed 3 times with PBS , bloc ked with 3% BSA in PBS , w ashed a gain with PBS, and incubated with appropriate primary antibodies for 16 h at 4 • C. Cells were washed with 0.1% PBST and incubated with the appropriate amount of anti-mouse MINUS and antira bbit PLUS pr obes supplied with Duolink in situ Red Starter Kit Mouse/Rabbit (DUO92101, Sigma-Aldrich, USA), according to the man ufactur er's instructions. Finall y, cov erslips wer e mounted onto slides using ProLong Gold Antifade Reagent containing DAPI and observation fields were randomly selected using confocal laser scanning microscope (LSM510; Zeiss, Oberkochen, Germany) equipped with 40x objecti v e and anal yzed using the Aim 4.2 softw ar e LSM510. The specificity of the signal was controlled by counting the dots produced when one of the primary antibodies was omitted. For counting the fluorescent dot signals, Ima geJ cell ima ge anal ysis softw ar e w as used. Relati v e n umber of dots/cell were plotted using Prism 9.4.1 softw ar e (GraphPad, USA).

Pr otein-pr otein Macr omolecular Docking
To investigate the possible interaction interface between RhoBTB1 and CUL3, the predicted structures of RhoBTB1 (UniProt ID: Q9DAK3) and CUL3 (UniProt ID: Q13618) were r etriev ed fr om AlphaFold ( https://alphafold.ebi.ac.uk/ ). B1B2C domain was extracted from full-length RhoBTB1 by removing the sequence specific to GTPase domain using PyMOL softw ar e (Schr ödinger, LLC). The models were docked using the ClusPro 2.0 webserver ( https://cluspro.org/ ). 23 This protein docking algorithm uses the fast Fourier transform corr elation appr oach combined with an automatic clustering method to propose interacti v e surfaces with fav ora b le fr ee energies and outputs 10 pr otein-pr otein docking states. 24 Pr otein-pr otein interface areas and molecular interactions were analyzed by PDBsum. 25 , 26 The inter action distances betw een amino acid pairs and images of interacting amino acids were generated after structural analyses of the protein files using UCSF Chimera software ( https: //www.rbvi.ucsf.edu/chimera ). 27

Pr o ximity Labeling By APEX2
We followed a typical APEX2 protocol as previously described. 28 Unless otherwise indicated, cells wer e tr eated with 1 μm MG132 for 16 h prior to harvesting to inhibit proteasomal de gr adation. Immediately prior to harvesting, H 2 O 2 (H1009, Sigma-Aldrich, USA) was added to a final concentration of 1 m m , the plates wer e gentl y a gitated, and incubated for exactl y 1 min at r oom temper ature . The reaction was quenched by washing 3 times with ice-cold Dulbecco's PBS containing 5 m m Trolox (238813, Sigma-Aldrich, USA), 10 m m sodium ascorbate (PHR1279, Sigma-Aldrich, USA), and 10 m m sodium azide (190385000, Thermo Fisher Scientific, USA). Cells were then lysed in RIPA lysis buffer supplemented with 1 m m PMSF, protease inhibitor cocktail (1861281, Thermo Fisher Scientific, USA), 5 m m Trolox, 10 m m sodium ascorbate, and 10 m m sodium azide, and then centrifuged at 15 000 × g for 10 min at 4 • C. Biotinylated protein was purified by incubation with str e ptavidin beads (88817, Thermo Fisher Scientific, USA) overnight with gentle rotation at 4 • C. The following buffers were used to thor oughl y w ash the beads to r emov e nonspecific binders: twice with RIPA lysis buffer, once with 1.0 M KCl (P41025, RPI Resear c h Products Int., USA), once with 0.1 M Na 2 CO 3 (L13098.36 Thermo Fisher Scientific, USA), once with 2.0 M urea (15505035, Thermo Fisher Scientific, USA) in 10 m m Tris (pH 8.0; T5941, Sigma-Aldrich, USA), and twice with RIPA lysis buffer. To examine the activity of the APEX2 tagged fusion construct using western blot, the biotinylated proteins were eluted by boiling the beads in 3 × protein loading buffer containing 20 m m DTT (D9779, Sigma-Aldrich, USA) and 2m m biotin (B4501, Sigma-Aldrich, USA) for 10 min, then r esolv ed on 10% SDS-PAGE and transferred to a nitrocellulose membr ane . Membr anes w er e b loc ked in 3% BSA in 0.1% Tw een-20 (P1379, Sigma-Aldrich, USA) for overnight at 4 • C. Immunoblotting was done using str e ptavidin HRP conjugate for 1 h at room temper ature . The membr ane w as w ashed with TBST a gain, and the biotinylation was visualized using enhanced chemiluminescence r ea gents according to the man ufactur er's instructions. For mass spectr ometr y, the suspension buffer w as r emov ed from the pulldown beads using a magnetic r ac k. Beads were resuspended in 100 μL of 40% Invitrosol (MS1000, Invitrogen, Thermo Fisher Scientific, USA) and 100 m m ammonium bicarbonate (09830, Fluka, Switzerland). Cysteines were reduced in 5 m m TCEP (646547, Sigma-Aldrich, USA) for 30 min at 37 • C and alkylated with 10 m m iodoacetamide (100351, MP Biomedicals, USA) for 30 min at 37 • C. Biotinylated proteome pulldown was digested dir ectl y on str e ptavidin beads ov ernight with 5 μg of tr ypsin (PR-V5113, Pr ome ga) at 37 • C follow ed by cleanup using the SP2 method. Thermo Scientific Pierce Peptide Retention Time Calibration Mixture (88321, Thermo Fisher Scientific, USA) was added at 4 n m concentration during dilution of the samples to 20 ng/ μL.

Mass Spectrometry Analysis
Each sample was analyzed on a Thermo Scientific Orbitrap Fusion Lumos MS via 3 technical r e plicate injections using a data-dependent acquisition (DDA) HCD MS2 instrument, using the parameters outlined in Supplementary Table S4. The technical r e plicates wer e b loc ked, and eac h bloc k was r andomized. Pooled QCs, which is a mixture of all the samples being anal yzed, wer e anal yzed at the start, end, and in betw een eac h sample. MS data wer e anal yzed using Pr oteome Discov er er 2.4 (Thermo) platform as outlined in the Supplementary Table S5. Protein identifications were filtered to include only those proteins identified by two or more unique peptides identified. Mass spectr ometr y pr oteomics data hav e been de posited to the Pr oteome Xchange Consortium via the PRIDE partner r e positor y with the dataset identifier PXD043245 and 10.6019/PXD043245.

Sta tistical Anal ysis
Results ar e expr essed as the mean of independent experiments ± SEM. The number of repeats ( n ) used in each experiment is indicated in the figures and text. Statistical significance was determined using one-way ANOVA using Prism 9.4.1 softw ar e (GraphPad, USA). A P -value less than 0.05 was considered statistically significant and indicated with an asterisk in figures.

PDE5 Interacts With B1B2C Domain of RhoBTB1
RhoBTB1 consists of five distinct domains: (a) an atypical GTPase domain, which may not exhibit GTPase activity, (b) a proline rich domain, (c) two BTB1 domains characteristic of those needed to associate with CUL3, and (d) a C-terminal domain ( Figure 1 A). RhoBTB1 is an adaptor protein designed to deliver specific substrates to the CUL3 ring ubiquitin ligase complex. PDE5 is the only known and validated substrate for RhoBTB1. To identify the minimal domain(s) on RhoBTB1 r equir ed to interact with PDE5, we generated a series of truncation mutants of RhoBTB1 carrying a single or multiple domains ( Figure 1 A). We performed a series of co-imm unopr ecipitation experiments in HEK293 cells transfected with Myc-tagged domains of RhoBTB1 and full length His-tagged PDE5. Truncated domains of RhoBTB1 wer e imm unopr ecipitated and anal yzed by western blotting for the presence of PDE5 in the immune complexes.
Single or combination domains, including GTPase, B1B2, PB1B2, B1, and B2 lacking the CT domain failed to interact with PDE5 despite evidence for their expression in lysates (Supplementary Fig. S1, data for B1 and B2 domains alone are not sho wn). Ho w ever, the addition of the CT domain to the tw o BTB domains (B1B2C) resulted in co-IP with PDE5 ( Figure 1 B). Recipr ocal co-imm unopr ecipitation anal ysis also confirmed the inter action betw een PDE5 and B1B2C ( F igur e 1 C). Nota b l y, the CT domain could not interact with PDE5 on its own ( Figure 1 D).
Collecti v el y, these findings suggested the CT domain is necessary but not sufficient to mediate binding to PDE5. To validate the specificity of the B1B2C:PDE5 interaction quantitati v el y, in situ PLA was performed. Consistent with the results obtained by co-imm unopr ecipitation, we obtained a strong PLA signal in the groups co-transfected with Myc-B1B2C and His-PDE5 ( Figure 2 ). No significant PLA signal was detected in the negative control groups either stained without primary antibody or transfected with RhoBTB1 without the CT domain (PB1B2).

Ectopically Expressed B1B2C Negati v ely Regulates Intracellular PDE5
A previous study from our group showed that RhoBTB1 acts as a substrate adaptor that deli v ers PDE5 to the CRL3 complex for its de gr adation, and thus, RhoBTB1 contr ols v ascular tone by regulating nitric oxide/cGMP signaling. 15 To determine whether B1B2C, the minimal r egion r equir ed to bind to PDE5, is also sufficient to regulate intracellular PDE5 levels, we transiently expr essed incr easing concentrations of B1B2C in HEK293 cells. The r elati v e lev el of His-ta gged PDE5 pr otein expr ession nota b l y decreased as the amount of B1B2C increased in the presence of cycloheximide, an inhibitor of protein synthesis ( Figure 3 A  and B). Also, in the presence of cycloheximide, the decrease in PDE5 caused by B1B2C ov er expr ession w as b locked by the pr oteasomal inhibitor MG132 suggesting a role for the proteasome ( Figure 3 C). Similarly, the B1B2C-mediated decrease in PDE5 was blocked by MLN4924, a neddylation inhibitor that acts as a pan-Cullin inhibitor. There was no decrease in PDE5 protein in presence of PB1B2 domain lacking the CT domain ( Supplementary  Figur e S2). Importantl y, ther e w as no change in PDE5 mRNA in the presence or absence of the B1B2C domain, consistent with a post-translation mechanism ( Figure 3 D). Thus, we concluded that the B1B2C domains are sufficient on their own to regulate the intracellular levels of PDE5, and that the GTPase and Prolinerich domains are dispensable.

B1B2C Promotes Pol yubiquitina tion of PDE5
BTB-domain containing proteins can interact and promote the polyubiquitination of their substrates through the CRL3 complex. 29 Ther efor e , w e asked if the decrease in PDE5 mediated by B1B2C occurs through ubiquitination of PDE5. His-tagged PDE5 was co-transfected with ubiquitin with either empty vector or Myc-tagged B1B2C in HEK293 cells. Imm unob lot anal ysis r ev ealed the extensi v e ubiquitination of PDE5 only in the presence of B1B2C ( Figure 3 E). Consistent with a role for CUL3, ubiquitination of PDE5 was blocked by MLN4914. To further delineate the role of CUL3 in B1B2C aided ubiquitination of PDE5, we expressed PDE5 and ubiquitin with either empty vector or B1B2C in a previously established CUL3-deficient HEK293 cell line made by CRISPR/Cas-9. In the absence of CUL3, PDE5 ubiquitination was drastically decreased in cells transfected with B1B2C. ( Figure 3 F). Interestingly, the ubiquitination of PDE5 was not completel y a bolished in CUL3 knockout HEK293 cells suggesting the possible involvement of alternati v e ubiquitination enzymes. Taken together, we concluded that B1B2C is r esponsib le for ubiquitination of PDE5 followed by its de gr adation via CUL3-dependent ubiquitin-proteasomal pathway, and the N-terminal GTPase and Proline-rich domains are dispensa b le.

Pr otein-pr otein Interaction Revealed CUL3 Binding Hotspots on B1B2C
Since the r esults a bov e showed that both CUL3 and the B1B2C domain of RhoBTB1 were critical for PDE5 ubiquitination, we studied the specific amino acids in B1B2C r equir ed for its binding to CUL3. We performed macr omolecular pr otein-pr otein doc king betw een CUL3 and B1B2C. Following pr otein-pr otein doc king, the doc king poses w er e filter ed based on cluster with maxim um n umbers to visualize the interface between CUL3 and B1B2C. Because PDBsum identifies pr otein-pr otein interactions as c hain-c hain inter actions, w e annotated the CUL3 and B1B2C domains as chains A and B, r especti v el y. 30 A PDBsum analysis of CUL3 and B1B2C reported an interacting interface comprising of 24 amino acid residues and the potential distance between them (Supplementary Figure S3). B1B2C-CUL3 complex formation is mediated by hydrogen bonds, electrostatic, and non-bonded interactions (Supplementary Table  S6). Further structural inv estigation r ev ealed that Pr o 353 (n umbering based on nati v e RhoBTB1) on B1B2C interacts with Arg 546 on CUL3 with the help of hydrogen bond and nonbonded interactions. Similarly, the interaction between Ser 363 on B1B2C and Arg 529 on CUL3 occurs b y h ydrogen bond and non-bonded interactions. In addition, interactions were also noticed for Ser 363 on B1B2C and Cys 522 , Asn 523 , and Ile 524 on CUL3.
We next tested whether the specific interacting residues identified in the CUL3-B1B2C complex are the major determinants of binding specificity using site directed mutagenesis and co-imm unopr ecipitation assays. We hypothesized that substituting the CUL3 binding residues on B1B2C would hamper its binding with CUL3, and ther efor e CUL3-de pendent de gr adation of PDE5. Results of co-imm unopr ecipitation r ev ealed that the interaction of B1B2C mutants Pro353Ala and Ser363Ala with CUL3 was markedly decreased compared to wild-type B1B2C and CUL3 despite exhibiting similar levels of expression in the l ysates ( Figur e 4 A and B). Inter estingl y, despite the decr eased binding of the mutants to CUL3, their binding to PDE5 was preserv ed ( Figur e 4 C). Consistent with decr eased binding to CUL3, both mutants were unable to mediate proteasomal de gr adation of PDE5 ( Figure 4 D and E, Supplementary Figure S4). Therefore, multiple sites along the interface are required for stabilization of the B1B2C and CUL3 complex.

Ascorbate Per o xidase 2 (APEX2) La beling Resolv es B1B2C Interacting Proteome
Next, we identified other interacting partners in the RhoBTB1-CUL3 pathway. We employed the APEX2 labeling system utilizing an APEX2-tagged B1B2C (B1B2C-APEX2) fusion construct. The outline of the experiment is depicted in Supplementary  Figure S5A. First, we showed that B1B2C-APEX2 was expressed in HEK293 cells as imm unob lotting confirmed the expression of the fusion construct at the expected molecular weight (Supplementar y Figur e S5B). Second, HEK293 cells transfected with B1B2C-APEX2 were subjected to biotin ylation in v olving tr eatment of cells with biotin-phenol and a fractional pulse of H 2 O 2 . Imm unob lotting with str e ptavidin-HRP conjugate indicated robust biotinylation of numerous proteins in the presence of the B1B2C-APEX2 construct, biotin-phenol, and H 2 O 2 demonstrating that the construct is functional (Supplementary Figure S5C). Consistent with this, no biotinylation was detected in non-transfected cells treated with either biotin-phenol or H 2 O 2 or both, or in transfected cells treated individually with of biotin-phenol or H 2 O 2 . 31 To validate the interaction capability of B1B2C-APEX2, we co-expressed B1B2C-APEX2 (as a Myctagged construct) and PDE5 in HEK293 cells. PDE5 was immunoprecipitated and analyzed by western blotting for the presence of B1B2C-APEX2 in the imm unopr ecipitate. Results of coimm unopr ecipitation r ev ealed that the interaction of B1B2C-APEX2 with PDE5 was preserved, validating that the APEX2 tag does not interfere with the interaction with PDE5 (Supplementar y Figur e S5D).
To rule out the surplus labeled proteome, we performed a contemporaneous screen using APEX2-tagged B1B2 (lacking the C terminal region) in HEK293 cells. Thus, comparing proteins identified by B1B2C-APEX2 versus B1B2-APEX2 allowed us to specificall y captur e the pr oteome interacting with C-terminal region of RhoBTB1. The biotinylated proteome was enriched using str e pta vidin beads and identified b y mass spectrometr y. Ov erall, out of a total of 2448 enriched pr oteins acr oss 6 samples (3 technical r e plicates acr oss two conditions: B1B2 and B1B2C), 268 proteins were selected based on statistical significance (Benjamini-Hochberg P -value < 0.05, Figure 5 A and Supplementar y Ta b le S7). Then, we selected 20 candidate proteins based on fold enrichment ( Figure 5 B). These proteins were found to be enriched more with B1B2C than B1B2 (abundance ratio B1B2C/B1B2: > 1.5-fold). A volcano plot graphically r e pr esents those proteins found to interact better with B1B2C (green), B1B2 (red), or below the level of statistical significance (blac k, F igure 5 C). Proteins chosen for further analysis are shown in blue.
These differ entiall y interacting pr oteins wer e further analyzed for their molecular processes and enriched pathways using the ShinyGO tool (version 0.76.3). In ShinyGO analysis, two pathw ays (nodes) ar e connected if they shar e 20% (default) or more genes. Darker nodes are more significantly enriched gene sets. Bigger nodes r e pr esent larg er g ene sets while thicker edges r e pr esent mor e ov erlapped genes. Most of the regulated pathways in our data set were related to calcium signaling, muscle contraction, nucleotide binding, and cytoskeletal remodeling (Supplementar y Figur e S6). This is consistent with our previous r e port that RhoBTB1 r egulated arterial stiffness thr ough the c ytoskeleton or ganization. 16

Physical Interactions of B1B2C Interacting Proteome Were Confirmed By Co-IP
We next validated the interaction of several of the top scored interactors detected by the pr oximity-de pendent la beling approach: SET Domain Containing 2 (SETD2), Calmodulin Regulated Spectrin Associated Protein 1 (CAMSAP1), and Tr affic king Protein Particle Complex Subunit 9 (TRAPPC9) for their physical interaction with B1B2C and B1B2 using Co-IP. We found each of these proteins interacted with B1B2C better than B1B2 ( Figure  6 A). Similarly, two other candidates (Annexin A6, ANXA6 and Calcium/calmodulin-de pendent Pr otein Kinase Kinase 2, CAMKK2) also interacted with B1B2C (data not shown). As an internal control, we selected High Mobility Group Box 1 (HMGB1), whic h inter acted better in the APEX2 screen with B1B2 when compared to B1B2C (B1B2/B1B2C abundance ratio: 1.57 versus 0.44 and P -value = 0.039). Consistent with the mass spectrometry data, HMGB1 was found to interact with B1B2 greater than B1B2C ( Figur e 6 B). Ther efor e, the high throughput mass spectr ometr y data corr oborated our co-imm unopr ecipitation findings. Gi v en this, we reasoned that SETD2, ANXA6, C AMKK2, C AMSAP1, and TRAPPC9 may interact with RhoBTB1.

Cellular Le v el of SETD2 Is Regulated By RhoBTB1 and CUL3
Evidence fr om pr evious studies suggested that SETD2 is de gr aded by the proteasomal pathway. 32 , 33 Given this, we hypothesized that the cellular levels of SETD2 might be controlled by RhoBTB1 and CUL3. First, we showed that the level of SETD2 protein was markedly increased in response to proteasomal (MG132) or Cullin (MLN4914) inhibition ( Figure 7 A). Second, the level of SETD2 was markedly increased in HEK293 CUL3KO cells, suggesting its regulation is through a CUL3-dependent mechanism ( Figure 7 B). Third, we used a siRN A approac h to knoc kdown the expression of RhoBTB1 in HEK293 cells. The knockdown of RhoBTB1 mRNA was variable with single siRNAs knocking down RhoBTB1 mRN A betw een 40%-80%, and combination or two or more siRNAs by about 70%-90% (data not shown). Whereas the effect on SETD2 mRNA was 2-fold or less (generally decreased),   the level of SETD2 protein (on av era ge) incr eased as RhoBTB1 decr eased ( Figur e 7 C). Taken together, our data suggested that SETD2 pr otein expr ession w as r egulated by RhoBTB1 and CUL3.

Discussion
We pr eviousl y r e ported that RhoBTB1 is a PPAR γ targ et g ene inv olv ed in r egulating arterial BP, v ascular function, and arterial stiffness. 12 , 16 One of the mec hanisms by whic h this occurs is that RhoBTB1 serves as a substrate adaptor for CUL3-mediated ubiquitination and proteasomal de gr adation of PDE5. 15 The CUL3-RhoBTB1-PDE5 r egulator y circuit is critical for the regulation of vasoconstriction and vasodilation because PDE5 controls the level of cGMP in vascular smooth muscle, the downstream mediator of endothelium-deri v ed nitric oxide. 6 , 18 We hypothesized that RhoBTB1 would exhibit a range of substrates because: (a) other BTB-domain containing proteins act as adaptors for CUL3-mediated regulation, 34 (b) other BTB-domain containing pr oteins hav e m ultiple substrates, 35 and (c) RhoBTB1 is ubiquitousl y expr essed. 36 We further hypothesized that the binding site for PDE5 on RhoBTB1 would be similar to the binding sites for other substrates on RhoBTB1. We ther efor e sought to identify the minimal RhoBTB1:PDE5 interaction domain to employ it as a bait to capture and identify other RhoBTB1-CUL3 substrates. Thus, this investigation made three fundamental discoveries regarding the biology of RhoBTB1. First, we demonstrated that the B1B2C domain of RhoBTB1 is the minimal binding region for PDE5. Second, we identified key residues required for the interaction between RhoBTB1 and CUL3, which are dispensable for the binding of RhoBTB1 to PDE5 but are required for PDE5 turnover. Third, using B1B2C as bait in APEX2-based proximity ligation, we identified interacting partners of RhoBTB1 and v alidated sev eral of them by co-imm unopr ecipitation. We further identified that one of them, SETD2, is regulated by the RhoBTB1-CUL3 pathway.
F irst, w e demonstr ated that RhoBTB1 binds to PDE5 utilizing its B1B2C domain. Inter estingl y, the B1B2C domain was sufficient, in the absence of the N-terminal GTPase and prolinerich domains, to deli v er PDE5 to the CRL3 complex for proteasomal de gr adation. This is interesting on its own as it suggests that the N-terminal portion of the protein is dispensa b le for its role as a substrate adaptor for PDE5 degradation, and perhaps other substrates for CUL3-mediated ubiquitination. Mor eov er, this suggests that RhoBTB1 may be a multifunctional protein, which r equir es its N-termin us for other functions. RhoBTB1 is a member of a family of atypical GTPases, but contains mutations in key residues, which may abolish or limit its GTPase activity, although this remains controversial. 37 , 38 Like other proteins with a proline-rich domain, RhoBTB1 ma y pla y a role in signal transduction or regulation of the cytoskeleton. 39 Indeed, w e show ed that RhoBTB1 may re gulate , in a manner which may or may not r equir e its r ole in the CRL3 complex, the state of actin pol ymerization in v ascular smooth m uscle cells. 16 RhoBTB1 has also been r e ported to interact with Rho kinases to inhibit cellular invasion. 40 The interaction between RhoBTB1 and Rho kinases has implications for vascular function. RhoBTB1 has also been r e ported to r e gulate the Golgi appar atus and bone resorption, among others. 18 , 41 , 42 Truncations of RhoBTB1 lacking the CT domain were unable to bind and pr oteol yticall y de gr ade PDE5. Thus, one of the important conclusions from this study is that the CT domain is essential for maintaining the functionality of RhoBTB1, at least with respect to ubiquitination and de gr adation of PDE5, but is not sufficient to bind to PDE5 on its own. Similarly, the CT domain was r equir ed to mediate an interaction between RhoBTB1, and other RhoBTB1-binding proteins identified by APEX2-mediated proximity ligation, including SETD2, CAMSAP1, TRAPPC9, ANXA6, and CAMKK2. In the case of SETD2, inhibition of either CUL3 or RhoBTB1 caused an increase in SETD2 abundance. One of the remaining questions is whether the interactions mediated by the N-terminal of RhoBTB1 coordinate with its C-terminal role as a substrate adaptor for CUL3. Inter estingl y, the famil y of RhoBTB pr oteins hav e been r e ported to exhibit autoinhibition by interaction of the N-terminus with the first BTB domain (BTB1), which becomes r eliev ed by binding to CUL3 and associated pr oteins. 35 CUL3 plays a key role by acting as a scaffold in the CRL3 complex, and hence, it aids in assembling the "substrate" proteins through BTB-domain containing proteins, such as RhoBTB1. Ther efor e, in the second phase of our study, we focused on deciphering the underlying molecular mechanisms inv olv ed in the binding of RhoBTB1 to CUL3. Macromolecular protein-protein docking r ev ealed that Pr o 353 and Ser 363 amino acid residues on B1B2C are crucial for its binding to CUL3. Nota b l y, point m utations of the CUL3 binding sites on the B1B2C domain of RhoBTB1 pr eserv ed normal binding to PDE5, but were unable to facilitate the proteasomal de gr adation of PDE5 despite exhibiting similar expr ession lev els. We ther efor e conclude that specific domains on RhoBTB1, CUL3, and PDE5 are all r equir ed for the formation and functionality of a pr oducti v e CRL3 complex regulating PDE5.
RhoBTB1 is transcriptionally regulated by PPAR γ . This is interesting because mutations in PPAR γ cause human HTN. 7 RhoBTB1 has also been identified in a genome wide association study to be associated with diastolic BP, and to be associated as an interacting loci in a study of over 1 million subjects. 43 , 44 Expression of RhoBTB1 is markedly decreased in mice carrying a transgene expressing dominant negative mutation in PPAR γ . 12 Expression of over 100 other PPAR γ -target genes were also decreased in these mice. Remarkably, the restoration of RhoBTB1 on its own by a PPAR γ -independent mechanism ameliorated adverse car dio vascular phenotypes in animals carrying a dominant-negative PPAR γ mutation. 15 RhoBTB1 is also downregulated in models of angiotensin-II mediated HTN and restoration of RhoBTB1 rapidly reverses arterial stiffness. 16 This data clearly implicated vascular smooth muscle PPAR γ -RhoBTB1-CUL3-PDE5 as an important regulator of arterial BP. However, it is notable that RhoBTB1 is ubiquitously expressed with the highest level of transcripts in skeletal muscle, kidney, placenta, and testis. 35 Single cell sequencing r ev eals detecta b le RhoBTB1 in a di v erse set of cell types, including adipocytes, macr opha ges, excitator y neur ons, ciliated cells of the respiratory tract, endometrial stromal cells, proximal tubular cells, placental syncytiotr ophob lasts, and a v ariety of v ascular cells types. We ther efor e m ust consider that like other BTB-domain containing proteins, which act as CUL3 adaptors, RhoBTB1 may hav e mor e than one substr ate , perhaps many.
To investigate this possibility, we chose to utilize a proximity labeling method that involved the use of APEX2-fusion proteins in HEK293 cells, followed by proteomic profiling with mass spectr ometr y. We took an appr oach to identify pr oteins, whic h inter acted with B1B2C greater than B1B2 (lacking the CT) to avoid identifying proteins, which may bind through the BTB domains, an ev olutionaril y conserv ed module , whic h is inv olv ed in pr otein-pr otein interactions. These domains ar e found in proteins that act as substrates for CUL3 but are also found in some zinc finger transcription factors and ion channels. 34 This was important as over 2400 interactions were identified in the screen, many of which overlapped between B1B2 and B1B2C. In r etr ospect, it might have been v alua b le to use B1B2C mutants, which pr ev ented interaction with CUL3 but pr eserv ed interaction with PDE5. However, the CUL3-RhoBTB1 experiments were performed concurr entl y with the APEX2 experiments. It remains unclear if that would pr ev ent binding to other BTB-domain containing proteins.
Gi v en that APEX2 labels not only direct interactors but also proteins in the proximity of the bait, w e w ere not surprised to see many potential targets. Indeed, we expected to observe many false positi v es. Howev er, the criteria of focusing on pr oteins that were detected with B1B2C greater than B1B2 allowed us to prioritize which proteins co-immunoprecipitated with B1B2C. Our validation studies also ruled out a proteome that might be reacting with the APEX2 tag itself. We also determined that HMGB1, which was detected in the APEX2 screen to bind to B1B2 greater than B1B2C, interacted in co-imm unopr ecipitation experiments with B1B2 better than B1B2C. Thus, the validation was successful in both directions.
Among the several RhoBTB1 binding partners deri v ed fr om the proteomic screen, w e c hose SETD2 to further investigate. SETD2 is a histone lysine methyltr ansfer ase inv olv ed in e pigenetic regulation, which may play a role in Huntington disease, Luscan-Lumish syndrome, Rabin-Pappas Syndrome, and Intellectual Developmental Disorder 70. [45][46][47] We focused on SETD2 because it was reported to regulate coronary vascular development in the mouse heart and may contribute to pulmonary HTN. 48 , 49 Indeed, we validated that SETD2 interacted with B1B2C and that its lev el of pr otein expr ession is markedl y enhanced: (a) by an inhibitor of Cullin proteins (MLN4924) or the proteasome (MG132), (b) in HEK293 cells with an CRISPR-Cas9-induced mutation in CUL3, and (c) by inhibition of RhoBTB1. Because of its high molecular w eight, w e had difficulty determining if it is dir ectl y ubiquitinated by RhoBTB1-CUL3, but all other evidence supports that hypothesis. Future studies will explore the role of SETD2 in vascular function and BP regulation. Indeed, given the success of the approach undertaken in HEK293 cells, we ar e curr entl y v alidating targets of RhoBTB1 interacting proteins identified in cultur ed smooth m uscle cells, and plan to employ the technique in placental cells where the level of RhoBTB1 expression is among the highest across tissues.

Ac kno wledgments
We wish to acknowledge and thank Dr Larry Agbor who generated the CUL3-deficient HEK293 cell line, Dr Suresh Kumar for assistance with microscopy, Dr Sonam Mittal for assistance with heatmaps and volcano plots, and Ms. Michaela Pereckas in the Center for Biomedical Mass Spectr ometr y Resear c h for assistance with identification of RhoBTB1 binding proteins. The gr aphic abstr act was created with BioRender.com under academic license through a departmental subscription to CDS.