Structural basis of interdomain communication in PPARγ

The nuclear receptor peroxisome proliferator-activated receptor gamma (PPARγ) regulates transcription via two activation function (AF) regulatory domains: a ligand-dependent AF-2 coregulator interaction surface within the C-terminal ligand-binding domain (LBD), and an N-terminal disordered AF-1 domain (NTD or A/B region) that functions through poorly understood structural mechanisms. Here, we show the PPARγ AF-1 contains an evolutionary conserved Trp-Pro motif that undergoes cis/trans isomerization, populating two long-lived conformations that participate in intradomain AF-1 and interdomain interactions including two surfaces in the C-terminal LBD (β-sheet and the AF-2 surface), which are predicted in AlphaFold 3 models but not AlphaFold 2. NMR and chemical crosslinking mass spectrometry show that interdomain interactions occur for soluble isolated AF-1 and LBD proteins, as well as in full-length PPARγ in a phase separated state. Mutation of the region containing the Trp-Pro motif, which abrogates cis/trans isomerization of this region, impacts LBD interaction and reduces basal PPARγ-mediated transcription and agonist-dependent activation of PPARγ. Our findings provide structural insight into published in vitro and cellular studies that reported interdomain functional communication between the PPARγ AF-1 and LBD suggesting some of these effects may be mediated via AF-1/LBD interactions.


INTRODUCTION
Peroxisome proliferator-activated receptor gamma (PPARγ; NR1C3) is a nuclear receptor transcription factor that controls gene expression programs in uencing the di erentiation of mesenchymal stem cells into adipocytes (adipogenesis), lipid metabolism, and insulin sensitivity. Like other nuclear receptors 1 , PPARγ is a multidomain protein that contains a central DNA-binding domain anked by two regulatory regions that in uence transcription: an N-terminal ligand-independent AF-1 domain (also called the NTD or A/B region) and a C-terminal ligand-dependent AF-2 coregulator interaction surface within the C-terminal LBD. e molecular and structural basis of ligand-regulated functions of the LBD of nuclear receptors are relatively well understood. For PPARγ, structural biology studies including X-ray crystallography, hydrogendeuterium mass spectrometry (HDX-MS), chemical crosslinking MS (XL-MS), and NMR spectroscopy have revealed how agonist ligands stabilize a transcriptionally active AF-2/helix 12 surface conformation upon binding to the orthosteric ligand-binding pocket in the LBD to promote coactivator protein recruitment and increased expression of PPARγ target genes that drive adipogenesis [2][3][4][5][6] . More recently, we reported a structural mechanism of ligand-dependent corepressorselective PPARγ inverse agonism 7 . We de ned the transcriptionally repressive AF-2/helix 12 conformation that promotes corepressor interaction and transcriptional repression of PPARγ. We also identi ed the apo-LBD conformation that dynamically exchanges between activeand repressive-like conformations 8 .
Despite these and other key advances in determining ligand-dependent structural mechanisms of nuclear receptor LBD function, the structural basis by which the disordered N-terminal AF-1 in uences the function of PPARγ and other nuclear receptors remains poorly understood. Only a few crystal structures of full-length nuclear receptors including PPARγ have been reported and these structures either lack electron density for the disordered AF-1 or the AF-1 was removed to facilitate crystallization [9][10][11][12] . Cryo-EM studies of nuclear receptors thus far have only provided low resolution (>10-25Å) structural snapshots of the AF-1 or, similar to the crystallography studies, protein samples were used where the AF-1 was removed [13][14][15][16] . Furthermore, AlphaFold 17 models of nuclear receptors frequently show cloud-like AF-1/NTD structural depictions that are thought to be an artifact of the computational method to avoid steric clashes with structured domains 18 .
Obtaining atomic resolution structural data on the PPARγ AF-1 would be important for the eld-new regulatory mechanisms are likely to emerge, and the data may explain published observations of interdomain functional communication between the PPARγ AF-1 and AF-2/LBD. For example, PPARγ transcription and PPARγ-mediated adipogenesis is increased by AF-1 removal, indicating the AF-1 negatively regulates PPARγ transcription 19 . Phosphorylation of Ser112 within the AF-1 negatively a ects LBD functions (ligand binding and coregulator interaction), downregulates the expression of PPARγ target genes, and inhibits adipogenesis [20][21][22][23] . Furthermore, phosphorylation of Ser112 in the AF-1 is inhibited by an agonist ligand binding to the LBD that stabilizes a transcriptionally active AF-2 surface conformation, but not by an inverse agonist that stabilizes a transcriptionally repressive LBD conformation 24 . ese observations suggest that the N-terminal AF-1 somehow alters the structure and function of the C-terminal LBD, and vice versa, though currently there is no structural evidence into the molecular mechanism.
To gain molecular insight into how the disordered AF-1 regulates PPARγ function, we used biophysical and structural biology approaches suitable for studying intrinsically disordered proteins (IDPs) including 1 Hand 13 C-detected NMR spectroscopy, single-molecule FRET (smFRET), and XL-MS. Our studies uncovered several previously unknown structural features of the AF-1 that are poised to regulate the structure and function of PPARγ. We uncovered a region of the AF-1 that natively exchanges between two long-lived structural conformations resulting from proline cis/trans isomerization at a Trp-Pro motif. Although the AF-1 is structurally disordered, our studies reveal intradomain contacts within the AF-1 indicating it adopts a partially compact conformational ensemble. Finally, we show that the LBD physically interacts with the AF-1, including the region containing the slowly exchanging Trp-Pro motif. Our ndings provide a structural mechanism to explain previous reports of interdomain functional communication. 11 , which likely explains why crystal structures that include full-length PPARγ do not show AF-1 electron density 11 ( Figure 1C). To provide further experimental support, we compared the isolated AF-1 and LBD using circular dichroism (CD) spectroscopy methods, which report on secondary structure features present in proteins 26 and temperaturedependent protein foldedness 27 . CD spectral analysis ( Figure 1D) of the AF-1 shows a random coil pro le characteristic of a disordered protein, whereas the LBD shows a pro le consistent with its three-layer α-helical sandwich fold 1 . CD collected as a function of temperature ( Figure 1E) shows a well-de ned unfolding transition for the LBD, whereas the AF-1 shows no notable cooperative unfolding transition.
Comparison of 2D [ 1 H, 15 N]-HSQC NMR spectra of the AF-1 and LBD ( Figure 1F) reveals the AF-1 displays poor amide spectral dispersion characteristic of a disordered protein, whereas the LBD-a structured three-layer α-helical sandwich fold that is stabilized upon ligand binding 6,28 -shows well dispersed amide chemical shi s. 15 N{ 1 H} heteronuclear NOE analysis reveals low values for the AF-1 ( Figure S1A) compared to the LBD (Figure S1B), indicating the AF-1 possesses higher amplitude motions on the ps-ns timescale, which is characteristic of IDPs. Analysis of backbone Cα and Cβ NMR chemical shi s, which can predict secondary structure in IDPs and folded proteins 29 , reveals AF-1 regions with low α-helical and β-sheet propensity ( Figure S1C) compared to the robust secondary structure present in the LBD ( Figure S1D). CD analysis of the AF-1 in the presence of increasing tri uoroethanol (TFE) shows an increase in α-helical signature ( Figure S1E). Taken together, these data indicate that although the AF-1 is structurally disordered, there are regions with transient or lowly populated secondary structure.

Charge repulsion influences AF-1 compactness
Disordered proteins can either adopt extended conformations with no intradomain contacts or compact conformational ensembles with transient or robust intradomain contacts-structural features that can be detected using paramagnetic relaxation enhancement (PRE) NMR methods 30 . Using site directed mutagenesis, we introduced a cysteine residue at several locations in the AF-1, which lacks native cysteine residues, and attached the cysteine-reactive nitroxide spin label MTSL to each construct (D11C, S22C, D33C, D61C, A91C, S112C). We collected 2D [ 1 H, 15 N]-HSQC NMR data in the paramagnetic and diamagnetic state and calculated peak intensity ratios (I PRE = I para /I dia ) to reveal AF-1 residues that are in close structural proximity to the MTSL label ( Figure  2A). Generally, residues with I PRE = 0 correspond to a distance <12Å, and I PRE > 0 and < 1 correspond to a distance between 13-25Å where the peak intensity decrease is proportional to 1/r 6 from the unpaired electron 31 . I PRE values of the MTSL spin label at D11 near the N-terminus (D11C-MTSL) are consistent with an extended conformation with no signi cant long-range contacts as the pro le is similar to a predicted I PRE pro le from an extended AF-1 ensemble calculated using exible-meccano 32 . In contrast, MTSL placement at other regions within the AF-1 reveals longrange contacts given the experimental I PRE values deviate from predicted extended I PRE AF-1 pro les. e experimental PRE NMR pro les indicate the most robust intradomain interactions occur for residues where the . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint MTSL is placed between D33 and S112, suggesting most of the AF-1 compact conformation occurs within this region. e sequence composition of the AF-1 shows a net negative charge ( Figure S2) and a calculated pI of 4.17, which led us to hypothesize negative charge-charge interactions may in uence the relative compactness of the AF-1 conformational ensemble. To test this, we performed smFRET using a two-color approach 33, 34 with an AF-1 double mutant construct in which cysteines were placed near the N-terminus (D33C) and C-terminus (Q121C). Increasing FRET e ciency (E FRET ) between the D33C and Q121C sites was observed at lower pH or increasing salt concentration ( Figure 2B), indicating that salt neutralization of the negatively charged AF-1 side chains results in a more compact AF-1 conformation. Saltdependent PRE NMR analysis of the AF-1 with the MTSL spin label placed at D33C validated these ndings since intramolecular PREs increased with increasing salt concentration ( Figure 2C). Taken together, the smFRET and PRE NMR data indicate the AF-1 transiently samples compact conformations that can be ne-tuned by changes in charge repulsion.

P40 influences AF-1 isomerization and inhibits transcription
While performing backbone NMR chemical shi assignment of the AF-1, we noticed a region displaying two populations of chemical shi s encompassing residues T34-N42 ( 34 TEMPFWPTN 42 ), a sequence conserved among both γ2 and γ1 isoforms and di erent PPARγ orthologs ( Figure 3A). is observation indicates this region of the AF-1 slowly interconverts between two structural populations on the millisecondto-second time scale, which can be detected using ZZ-exchange NMR methods. Only one tryptophan residue (W39) is present in the AF-1, but two tryptophan side-chain indole peaks are present with populations of ~85% and ~15% ( Figure 3B). ZZ-exchange measurements at 25°C revealed no cross-correlated exchange for these W39 indole peaks. However, ZZ-exchange measurements at elevated temperatures revealed an increasing population of cross-correlated peaks indicating this region of the AF-1 slowly interconverts or isomerizes between at least two structural conformations on a timescale (τ ex ) of milliseconds/seconds at elevated temperature to seconds/minutes at lower temperatures.
Two proline residues, P37 and P40, are located within this slowly isomerizing region, which led us to hypothesize that proline cis/trans isomerization contributes to the mechanism. Chemical shi values of proline Cβ and Cγ nuclei are predictive of cis or trans proline conformations 35,36 . We therefore analyzed CC(CO)NH NMR data of the AF-1 ( Figure 3C) and found that the Cβ and Cγ chemical shi s of the two P37 conformations populate a trans conformation. In contrast, the two P40 conformations showed Cβ and Cγ chemical shi s consistent with one trans and one cis conformation. ese data pinpoint the W39-P40 (Trp-Pro) motif, a dipeptide sequence previously shown to enrich the cis isomer 37 , as the likely origin of the slowly isomerizing AF-1 conformational switch.
To con rm the role of cis/trans isomerization of P40 in populating the two long-lived AF-1 conformations, we compared 2D [ 1 H, 15  was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint Structural basis of interdomain communication in PPARγ | bioRxiv | July 13 th , 2022 | 4 NMR of P40A and W39A AF-1 mutants to WT AF-1 ( Figure 3D). e P40A mutant suppresses but does not completely abolish cis/trans isomerization of this region, changing the relative cis/trans populations from ~85%/15% in WT to ~94%/6% ( Table S1). As expected, W39 indole peaks are absent in the W39A mutant; however, cis/trans populations are still observed in this mutant (~89%/11%) that are more similar to WT levels, suggesting W39 may be dispensable for cis/trans isomerization.
To determine how the mutants impact PPARγ transcription, we performed a cell-based transcriptional reporter assay ( Figure 3E) along with western blot analysis of protein levels ( Figure S3). HEK293T cells were transfected with a PPAR-responsive luciferase reporter plasmid along with expression constructs encoding wild type (WT) PPARγ or mutant variants (W39A or P40A) to test the functional role of the Trp-Pro motif. We also tested a DBD-hinge-LBD construct lacking AF-1 (ΔAF-1), which was previously shown to increase PPARγ transcription in di erentiated mouse adipocytes. In the absence of ligand, the P40A construct displayed an increase in transcription relative to WT PPARγ, while there was no signi cant e ect of ΔAF-1 or W39A constructs. When HEK293T cells were treated with the synthetic PPARγ agonist rosiglitazone to mimic the condition where endogenous ligands are produced in di erentiated adipocytes, the ΔAF-1 and P40A constructs displayed an increase in transcription relative to WT PPARγ, whereas the W39A mutant displayed no signi cant change. ese ndings indicate that removal of the AF-1 inhibits ligand-dependent activity, while the inhibitory e ect of P40 cis/trans isomerization is ligand-independent.
e minimal e ect of the W39A mutation on cis/trans populations by NMR and transcription in cells suggests that W39 may be dispensible for isomerization and that isomerization may be driven by alternative mechanisms.

AF-1/LBD interdomain interaction in full-length PPARγ
To determine whether the two long-lived AF-1 conformations occur in full-length PPARγ, we compared 1D [ 1 H]-NMR spectra of AF-1 and full-length PPARγ. Focusing on the indole peaks of W39 ( Figure 4A), which is the only tryptophan residue in PPARγ, two peaks are observed in full-length PPARγ that are shi ed down eld (i.e., to the le ) relative to the peaks observed in the AF-1 alone. Addition of the agonist ligand rosiglitazone, which stabilizes an active LBD conformation, shi s the full-length PPARγ W39 indole peaks further down eld, indicating that ligand binding at the LBD a ects the chemical environment of the AF-1.
To extend our 1D [ 1 H]-NMR ndings, we used sortase A-mediated protein ligation and segmental isotope labeling 38,39 to generate fulllength PPARγ protein where only the AF-1 is 15 N-labeled and visible in 2D [ 1 H, 15 N]-HSQC NMR data. Overlay of 2D NMR spectra of segmentally [ 15 N-labeled AF-1] full-length PPARγ construct and the isolated 15 N-labeled AF-1 reveals select chemical shi and line broadening changes ( Figure 4B). For example, in the full-length PPARγ data the peak corresponding to the last residue in the isolated AF-1 (M135) is not present and several other C-terminal residues show chemical shi changes, con rming the segmentally labeled sample observed for fulllength PPARγ is completely ligated. e two W39 indole peaks are shi ed down eld relative to the isolated AF-1 spectrum, consistent with the 1D NMR analysis of isolated AF-1 and full-length PPARγ. Furthermore, residues in AF-1 regions involved in intramolecular PRE contacts (Figure 2) also show chemical shi and peak line broadening changes in the di erential 2D NMR data. Taken together, these ndings indicate the  TEMPFWPTN  TEMPFWPTN  TEMPFWPTN  TEMPFWPTN  TEMPFWPTN  TEMPFWPTN  TEMPFWPTN   human  mouse  rat  bovine  pig  macaque  (A) Residues within the AF-1 displaying two populations of NMR peaks localize to a region containing two proline residues (P37, green; P40, pink), one of which is adjacent to a tryptophan residue (W39). (B) ZZ-exchange NMR data (delay = 2 s) focused on the W39 indole peaks shows cross-correlated peaks appearing at elevated temperatures that are missing at lower temperatures (dashed gray oval). (C) Strips from 3D CC(CO)NH NMR data focused on P37 (i-1 to F38 amides) and P40 (i-1 to T41 amides). Average chemical shift ranges for trans and cis confomers of proline Cγ and Cβ atoms are displayed in green (trans) and pink (cis). (D) Overlay of 2D [ 1 H, 15 N]-HSQC NMR spectra comparing wild-type AF-1 to P40A (green) or W39A (pink) mutants. (E) Cell-based luciferase reporter assay in HEK293T cells to assess the ability of full-length PPARγ and mutant constructs to activate transcription without or with treatment of the synthetic PPARγ agonist rosiglitazone (1 µM) using a 3xPPRE-luciferase reporter plasmid (n=3). Bars and error bars represent the mean and s.d.; data representative of >2 experiments.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint chemical environment of the AF-1 is di erent within the context of the full-length protein in a ligand-dependent manner. Furthermore, these data suggest there may be interdomain contacts in full-length PPARγ, such as a physical AF-1/LBD interaction, that involve regions of the AF-1 involved in intradomain contacts de ned by our PRE NMR analysis. To corroborate our NMR ndings that the AF-1 and LBD physically interact in full-length PPARγ, we performed chemical crosslinking mass spectrometry (XL-MS) comparing DSSO-crosslinked samples of full-length PPARγ to a truncated construct without the AF-1 containing the DBD-hinge-LBD (ΔAF-1). e di erential analysis revealed 38 crosslinks enriched in full-length PPARγ ( Figure 4C) including 16 unique intradomain AF-1 crosslinks and 13 unique AF-1/LBD crosslinks ( Figure 4D). Crosslinks involving the AF-1 localize primarily to the C-terminal half of the AF-1 between K94 and K125, which contains 5 out of 6 lysine residues present in the AF-1. Seven LBD residues involved in interdomain AF-1 crosslinks include residues near the β-sheet surface on helix 1 (K244, K252), helix 2 (K268), Ω-loop (K303), helix 6 (K382), and helix 10 (K462)-as well as a lysine residue near the AF-2 surface (K432).
ese LBD crosslinks are consistent with the putative location of the AF-1 region in the crystal structure of full-length PPARγ where electron density for the AF-1 was not observed (Figure 4E) 11 . Furthermore, 7 crosslinks are enriched in ΔAF-1 PPARγ, of which 6 are LBD/LBD   (D) Plot of the P-value vs. fold change of DSSO crosslinks from the differential XL-MS analysis. (E) LBD residues involved in AF-1 crosslinks plotted on the AlphaFold structure of PPARγ (residues 136-505); the putative location of the AF-1, which is missing in crystal structures of full-length PPARγ, is shown in pink.
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint crosslinks that also localize near the β-sheet surface-indicating the AF-1, and by corollary its removal, structurally a ects the LBD conformation.

Trp-Pro motif is a major determinant in LBD interaction
e XL-MS data of full-length PPARγ, though limited to detecting interactions to the C-terminal half of the AF-1, show this region interacts with the β-sheet LBD surface. However, the di erential AF-1 vs. sortaseligated full-length NMR data suggest a more extensive interaction involving other regions of the AF-1. We therefore performed NMR chemical shi structural footprinting analysis to more completely map the interaction between the isolated AF-1 and LBD. To map the LBD-binding interface within the AF-1, we collected 2D [ 1 H, 15 N]-HSQC data ( Figure  5A) and 2D [ 13 C, 15 N]-(HACA)CON data ( Figure 5B) of 15 N-or 13 C, 15 Nlabeled AF-1, respectively, in the absence and presence of LBD. Addition of LBD caused changes in AF-1 chemical shi s (fast exchange on the NMR time scale) and peak line broadening (intermediate exchange on the NMR time scale) for select residues, con rming a direct interaction between the isolated domains. ere are three notable features apparent in these NMR structural footprinting data.
First, AF-1 residues most a ected that display the largest chemical shi changes and peak line broadening include F38, W39, P40, and T41-pinpointing the slowly exchanging Trp-Pro motif as a major determinant in the LBD interaction. Similar to the 1D and 2D di erential NMR data of full-length PPARγ and sortase-ligated full-length PPARγ, the W39 indole side chain peaks are shi ed down eld with increasing LBD concentrations. Furthermore, chemical shi perturbations or peak movements for other backbone amide residues occur in a similar direction between the isolated AF-1/LBD titration data and the di erential sortase full-length NMR data.
ese observations indicate the interaction between the isolated AF-1 and LBD interaction is similar to what occurs within the context of full-length PPARγ.
Second, AF-1 residues distal in primary sequence from the Trp-Pro motif show chemical shi perturbations in the NMR structural footprinting data and correspond to regions involved in intramolecular PRE contacts to the Trp-Pro region. Chemical shi changes and peak  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint line broadening is also observed for these residues in the 2D di erential sortase NMR data on full-length PPARγ vs. isolated AF-1. ese observations suggest a more extensive LBD interaction interface or allosteric conformational changes that occur in the AF-1 upon LBD interaction. is di erence in exchange on the NMR time scale suggests that the cis AF-1 conformation may interact with the LBD with higher a nity and/or di erent interaction kinetics than the trans AF-1 conformation.
Trp-Pro AF-1 motif interacts with two LBD surfaces: the β-sheet and AF-2 To map the AF-1-binding interface within the LBD, we collected 2D [ 1 H, 15 N]-TROSY-HSQC data of 15 N-labeled PPARγ LBD in the absence and presence of AF-1. Apo/ligand-free PPARγ LBD dynamically exchanges between transcriptionally active and repressive conformations on the intermediate exchange NMR time scale resulting in missing or very broad peaks for about half of the LBD including residues within and near the orthosteric ligand-binding pocket and AF-2 surface 6, 28 due to dynamic exchange between active and repressive LBD conformations 8 . Upon binding a full agonist such as rosiglitazone, which stabilizes an active LBD conformation, nearly all NMR peaks are visible 6,8 . We therefore performed studies in the apo-LBD form and in the presence of rosiglitazone to fully map the AF-1 interaction surface on the LBD. Titration of AF-1 into 15 N-labeled apo-LBD revealed chemical shi perturbations for several NMR peaks (Figure 6A). Residues that show the largest chemical shi changes colocalize near the β-sheet surface and include R262, T266, and T269 on helix 2; D408 on the helix 7-8 loop; and H453 on the helix 9-10 loop. To obtain a more holistic view of the AF-1 binding surface, we titrated AF-1 into 15 N-labeled LBD bound to rosiglitazone and observed additional chemical shi perturbations in regions that are stabilized in the active conformation ( Figure 6B). Agonist-stabilized residues that show the largest chemical shi changes include additional residues on the β-sheet surface (F275, V276, I277, G286, R378); as well as residues within the AF-2 coregulator interaction surface (K347, Y348, and V350 on helix 5; and L496 on helix 12).
NMR chemical shi footprinting analysis reports on both residues directly involved in the AF-1/LBD interaction as well as allosteric conformational changes that occur upon binding. We therefore performed intermolecular PRE NMR to speci cally map where the AF-1 region containing the Trp-Pro motif interacts on the LBD. We collected 2D [ 1 H, 15 N]-TROSY-HSQC data of 15 N-labeled LBD bound to rosiglitazone in the presence of AF-1 with a MTSL spin label placed at D33 (D33C-MTSL) near the Trp-Pro motif ( Figure 7A). Di erential analysis of AF-1 D33C-MTSL in the paramagnetic and diamagnetic states showed decreases in NMR peak intensity ( Figure 7B) for residues at two distinct LBD surfaces ( Figure 7C): a large grouping of residues comprising the AF-2 coregulator interaction surface, and a smaller grouping of residues near the β-sheet surface.
To con rm the AF-2 surface interaction with the Trp-Pro motif, we took advantage of a double mutant LBD construct we previously generated (C313A/K502C) where we introduced a cysteine residue on the AF-2 surface within helix 12 (K502C) and removed the only cysteine residue in the LBD (C313A) to enable site-speci c covalent labeling to the AF-2 helix 12 8,40,41 . Using this construct, we performed intermolecular PRE NMR with 15 N-labeled AF-1 in the presence of K502C-MTSL LBD. Di erential analysis with apo K502C-MTSL LBD revealed the largest decrease in I PRE values within and near the Trp-Pro motif though some smaller PRE e ects are also observed where the I PRE values dip below 0.9 for residues within regions involved in intradomain AF-1 contacts ( Figure 7D). In the presence of rosiglitazone-bound K502C-MTSL LBD, compared to apo K502C-MTSL LBD, a larger decrease in peak intensity is observed for the W39 trans indole conformation but not the cis conformation ( Figure 7E). is could indicate an interaction preference with the Trp-Pro motif in the trans AF-1 conformation for agoniststabilized active LBD conformation.
Taken together, the intermolecular PRE data corroborate the NMR structural footprinting data and extend the XL-MS data revealing that both the β-sheet and AF-2 LBD surfaces are involved in the AF-1 interaction. Furthermore, our studies pinpoint the slowly exchanging AF-1 Trp-Pro motif, which populates two long-lived AF-1 structural conformations, as a major interaction determinant with the AF-2 coregulator interaction surface in the LBD.  . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint mechanism 19-23 . However, the structural basis for AF-1 activity, including its e ects on the LBD, has remained elusive due to its disordered properties that make it di cult to study by X-ray crystallography or cryo-EM. Here, we used a combination of solution-based techniques to characterize the structural properties of the PPARγ AF-1. We found that although predominantly disordered, the AF-1 is characterized by transient long-range intramolecular contacts that can be enhanced by decreasing negative charge repulsion. us, a change in AF-1 structure and compaction could occur under cellular states that would a ect protein net charge, such as the presence of binding partners including other proteins, nucleic acids (mRNA and chromatin), or small molecule metabolites 1 ; upon post-translational modi cation of PPARγ 42 ; or cellular compartments with localized changes in ionic strength or pH such as phase separated biomolecular condensates 43 .
Our NMR analysis also revealed that the AF-1 contains an evolutionarily conserved Trp-Pro motif that undergoes cis/trans isomerization on a slow timescale (milliseconds to seconds) populating two long-lived AF-1 conformations. Tryptophan is known to facilitate isomerization of neighboring proline residues 37 ; however the W39A mutant failed to fully disrupt isomerization and there was some residual isomerization even in the P40A mutant. Residues preceding the Trp-Pro motif, including P37 and F38, may also contribute to isomerization of this region. e strong propensity for isomerization despite these mutations suggests it is a robust dynamic property of this region, and the observation that the P40A mutation increases cell-based transcriptional activity of PPARγ indicates isomerization of this region may have important cellular functions that warrants further study.
Previous studies showed the AF-1 inhibits ligand-dependent PPARγ activity, but did not determine whether the e ects were due to a direct AF-1/LBD interaction [19][20][21][22][23] . However, AF-1/NTD interactions with nuclear receptor LBDs have been reported or suggested for at least four other nuclear receptors including androgen receptor 44-51 , estrogen receptor 52, 53 , glucocorticoid receptor 54 , and mineralocorticoid receptor 55, 56 . Using CSP and PRE NMR, we tested for AF-1/LBD binding and found the AF-1 contacts the LBD at two distinct surfaces. Unlike the AF-1/AF-2 interaction identi ed for androgen receptor, in which an FXXLF motif within the AF-1 folds upon binding, the "fuzzy" interaction observed for PPARγ aligns with others previously identi ed for disordered proteins 57,58 . Intriguingly, our results indicate the AF-2 surface of the LBD predominantly interacts with the region encompassing the AF-1 Trp-Pro motif, which notably is present in both major PPARγ isoforms (γ1 and γ2). Although our NMR data suggest the interaction may be somewhat weak when the isolated AF-1 and LBD are titrated together, given the chemical shi perturbations occur on the fast-to-intermediate NMR time scale, within the context of full-length PPARγ the interaction is likely to be more robust since the two domains are physically tethered together joined by the DBD and hinge region.
Previous studies indicated the AF-1 inhibits PPARγ function such that AF-1 deletion results in increased PPARγ transcription in di erentiated mouse adipocytes 19 . Our experiments using human HEK293 cells revealed the inhibitory e ect of the AF-1 on PPARγ transcription was only evident in the presence of an activating PPARγ ligand. Although the prior study did not add a PPARγ ligand, mouse adipocytes are known to produce activating PPARγ ligands naturally during adipogenesis 59 . us, it will be important to explore how AF-1 binding to the LBD inhibits ligand binding or a ects coregulator binding at the AF-2 surface. Given our data that the AF-1 interacts with the AF-2 surface, the AF-1 could directly compete for coactivator binding as previously observed for the . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint AR and ER AF-1 51, 53 , which would explain why AF-1 deletion increases PPARγ transcription. Likewise, AF-1 interactions at the LBD β-sheet surface could interfere with ligand exchange into the orthosteric ligandbinding pocket 4 , conformational changes associated with ligand binding, or interaction with other nuclear receptors 11 or other proteins. Proline cis/trans isomerization plays important roles in protein folding 60 and in the function of proteins with examples including isomerizationregulated ligand-gated ion channel pore opening 61 and isomerization within a linker region that controls interdomain autoinhibitory interactions [62][63][64] . More recently, studies have revealed functional regulatory roles for proline cis/trans isomerization in IDPs including regulation of protein-protein interactions 65,66 , conformation-speci c phosphatase enzymatic activity towards phosphorylated serine-proline motifs 67 , misfolding/aggregation of IDPs 68 , and circadian transcriptional regulation 69 . Our discovery of the PPARγ AF-1 Trp-Pro motif interaction with the AF-2/LBD is, to our knowledge, a unique example of a disordered domain undergoing cis/trans isomerization that participates in an interdomain interaction with a structured domain. Since the LBD interaction occurs at the site of AF-1 cis/trans isomerization, it will be important in future studies to determine if AF-1 isomerization in uences LBD binding and activity. Given our NMR data, which indicate the LBD interaction with the cis and trans AF-1 conformations may occur with di erent kinetic exchange properties and the LBD may preferentially interact with the trans isomer in the presence of activating ligand, changes to the isomerization rate or relative isomer populations (i.e., by an isomerase enzyme) could also modulate the AF-1/LBD interaction.
In addition to characterizing a direct interdomain interaction between the PPARγ AF-1 and LBD, our solution-based structural approach illuminates a platform to interrogate the activities of disordered nuclear receptor domains. While nearly all of the 48 human nuclear receptors contain unstructured AF-1 regions with important documented functional roles, the structural basis for their activities remain largely unexplored. us, it will be important to apply structural approaches such as those outlined here to gain a more comprehensive understanding of nuclear receptor activity.

Protein expression and purification
Proteins were expressed in Escherichia coli BL21(DE3) cells (Life Technologies). Full-length and ΔAF-1 PPARγ proteins were expressed using an autoinduction procedure where cells were grown for 6 hrs at 37 ˚C followed by addition of 0.1 mM zinc chloride, temperature reduction to 22 ˚C for 16 hrs. PPARγ AF-1, LBD, and sortase A-compatible proteins were either expressed in Terri c Broth (TB) media or M9 media supplemented with 13 C-glucose and/or 15 NH 4 Cl (Cambridge Isotope Labs, Inc.) followed by protein induction with 0.5 mM isopropyl ß-D-1-thiogalactopyranoside (Gold Biotechnology) at 18˚C for 16 hrs. Cells were harvested by centrifugation, washed with phosphate bu ered saline (PBS) bu er, and resuspended in a cell lysis bu er containing 40 mM potassium phosphate (pH 7.4), 500 mM potassium chloride, 15 mM imidazole, 0.5mM ethylenediaminetetraacetic acid (EDTA), 1 mM phenylmethylsulfonyl uoride (PMSF), and Pierce protease inhibitor tablets ( ermo Scienti c). Cells were lysed by sonication and the lysate was clari ed by centrifugation at 14000 rpm for 45 min and ltered with a 0.2µm lter prior to loading into the Ni-NTA column. e protein was eluted against a 500 mM imidazole gradient through a Ni-NTA column, followed by overnight dialysis against a bu er without imidazole for TEV protease His tag cleavage at 4°C. e next morning, the sample was loaded onto the Ni-NTA column for contaminant and tag removal. e ow through containing the puri ed protein was collected, concentrated, and ran through an S75 size exclusion column (GE healthcare) in the NMR bu er (20 mM KPO 4 pH 7.4, 50 mM KCl, and 0.5 mM EDTA. For PPARγ full-length and LBD the corresponding protein peak was collected and stored at -80 ˚C prior to use. For PPARγ AF1 the corresponding protein peak was collected, concentrated to 5mL and boiled at 70˚C for 10 min. e sample was centrifuged at 4000 rpm for 15 min, ltered with a 0.2µm lter and loaded into a Q column (GE healthcare) to remove suspected lingering proteases and DnaK. e column was eluted with a 1M potassium chloride gradient where the most prominent peak corresponds to our protein of interest. e peak was collected and dialysed overnight in the NMR bu er. e next day, the protein was concentrated and stored at -80˚C. All puri ed proteins were veri ed by SDS-PAGE as >95% pure and 1mM TCEP was added to all the bu ers during PPARγ full-length puri cation.

CD spectroscopy
Circular dichroism (CD) data of PPARγ AF-1 and LBD were collected using a Jasco J-815 CD Spectropolarimeter using a bu er containing 10 mM potassium phosphate and 50 mM potassium uoride. An average of 3 scans were recorded per measurement using a scan speed of 100 nm/ min at room temperature (~23 ˚C) between a spectral range of 190-260 nm using a 1-mm optical bandwidth. ermal unfolding curves were obtained by increasing the temperature from 0-00 ˚C with measurements recorded at 222 nm and 200 nm.

MTSL nitroxide spin labeling for PRE NMR studies
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint MTSL labeling of proteins was performed following a published protocol 31 . Brie y, proteins were concentrated to ~300-500 µM in 1 mL, reduced with 1.25 mM dithiothreitol (DTT), and passed through a Zeba desalting column ( ermo Scienti c) that had been equilibrated with NMR bu er; the eluate was collected, wrapped in aluminum foil, and supplemented with NMR bu er containing 10 molar equivalents of MTSL (Cayman Chemical #16463; also called MTSSL)from a stock solution in 100% DMSO-d 6 . e labeling reaction was set up at room temperature (~23 ˚C) for 15 min with gentle mixing followed by an overnight (~18 hrs) incubation a er addition of another 10 molar equivalents of MTSL. e following day, the MTSL-labeled protein was concentrated to 400 µL and dialysed for 16 hrs at 4 ˚C in the dark to remove unconjugated MTSL.

NMR spectroscopy
NMR experiments were acquired at 298 K (unless indicated otherwise) on a Bruker 700 MHz NMR instrument equipped with a QCI-P cryoprobe or a Bruker 600 MHz NMR instrument equipped with a TCI-H/F cryoprobe. NMR samples were prepared in NMR bu er (50 mM potassium phosphate, 20 mM potassium chloride, 1 mM TCEP, pH 7.4, 10% D 2 O) and typically contained 100-600 µM of the isotopically labeled component ( 13 C and/or 15 N); no signi cant chemical shi di erences were observed within this concentration range. 3D NMR experiments used for AF-1 chemical shi assignment were collected using a 300 µM 13 C, 15  were calculated from the ratio of peak intensities (I para /I dia ) as previously described 31 . NMR data were collected using Bruker Topspin (v3.2) so ware, and processed/analyzed using NMRFx and NMRViewJ 74,75 .

DSSO chemical crosslinking of protein samples
Full-length PPARγ and PPARγ ΔAF-1 (residues 136-505) were diluted to 10 μM in a bu er containing 20 mM potassium phosphate (pH 8.0), 150 mM KCl, 0.5 mM sodium citrate, and 10% glycerol. DSSO crosslinker ( ermoFisher #A33545) was freshly dissolved in DMSO to a nal concentration of 75 mM and added to the protein solution at a nal concentration of 1.5 mM. e reaction was incubated at 25 °C for 1 hr and then quenched by adding Tris bu er (pH 8.0) to a nal concentration of 50 mM and incubating an additional 10 min at 25 °C. Control reactions were performed in parallel without adding the DSSO crosslinker. All crosslinking reactions were carried out in three replicates. Crosslinked samples were con rmed by SDS-PAGE and Coomassie staining along with negative control samples that were not treated with DSSO. Samples were separately pooled, precipitated using acetone, and dried protein pellets were resuspended to 12.5 μL in a bu er containing 50 mM ammonium bicarbonate (pH 8.0) and 8 M urea. ProteaseMAX (Promega, V5111) was added to the resuspended samples to a nal concentration of 0.03%, the solutions were mixed on an orbital shaker operating at 1000 rpm for 15 min, and then 87.5 μL of 50 mM ammonium bicarbonate (pH 8.0) was added. Samples were digested for 4.5 hrs using trypsin added at a ratio of 1:170 (w/w trypsin:protein) at 37 °C then subsequently digested for 18 hrs using chymotrypsin at a ratio of 1:85 (w/w chymotrypsin:protein) at 25 °C. e resulting peptides were acidi ed to 0.67% tri uoroacetic acid (TFA) and then desalted using C18ZipTip (Millipore cat no. ZTC18 5096). Dried peptides were frozen, stored at -20°C, and resuspended in 10 μL of 0.1% TFA in water prior to LC-MS analysis.
Chemical crosslinking mass spectrometry (XL-MS) 500 ng of sample was injected (triplicate injections for both control and crosslinked samples) onto an UltiMate 3000 UHP liquid chromatography system (Dionex, ermoFisher). Peptides were trapped using a μPAC C18 trapping column (PharmaFluidics) using a load pump operating at 20 μL/min. Peptides were separated on a 200 cm μPAC C18 column (PharmaFluidics) using the following gradient: 5% Solvent B for 70 min, 30% Solvent B from 70 to 90 min, 55% Solvent B from 90 to 112 min, 97% Solvent B from 112 to 122 min, and 5% Solvent B from 120 to 140 min, at a ow rate of 800 nL/min. Gradient Solvent A contained 0.1% formic acid, and Solvent B contained 80% acetonitrile and 0.1% formic acid. Liquid chromatography eluate was interfaced to an Orbitrap Fusion Lumos Tribrid mass spectrometer ( ermoFisher) through a Nanospray Flex ion source ( ermoFisher). e source voltage was set to 2.5 kV, and the S-Lens RF level was set to 30%. Crosslinks were identi ed using a previously described MS2-MS3 method with slight modi cations 76 . Full scans were recorded from m/z 350 to 1,500 at a resolution of 60,000 in the Orbitrap mass analyzer. e AGC target value was set to 4×10 5 , and the maximum injection time was set to 50 ms in the Orbitrap. MS2 scans were recorded at a resolution of 30,000 in the Orbitrap mass analyzer. Only precursors with a charge state between 3 and 8 were selected for MS2 scans. e AGC target was set to 5×10 4 , a maximum injection time of 54 ms, and an isolation width of 1.6 m/z. e CID fragmentation energy was set to 25%. e two most abundant reporter doublets from the MS2 scans with a charge state of 2-6, a 31.9721 Da mass di erence 77 , and a mass tolerance of ±10 ppm were selected for MS3. e MS3 scans were recorded in the ion trap in rapid mode using HCD fragmentation with 35% collision energy. e AGC target was set to 20,000, and the maximum injection time was set for 150 ms and the isolation width to 2.0 m/z.
To identify crosslinked peptides, ermo .Raw les were imported into Proteome Discoverer 2.5 ( ermoFisher) and analyzed via the XlinkX algorithm 78 using the MS2_MS3 work ow with the following parameters: MS1 mass tolerance, 10 ppm; MS2 mass tolerance, 20 ppm; MS3 mass tolerance, 0.6 Da; digestion, trypsin-chymotrypsin with ten missed cleavages allowed; minimum peptide length of ve amino acids; and DSSO (K, S, T, Y). e XlinkX/PD Validator node was used for crosslinked peptide validation with a 5% false discovery rate (FDR). Identi ed crosslinks were further validated and quanti ed using Skyline (version 21.1) 79 using a previously described protocol 80 . Crosslink spectral matches found in Proteome Discoverer were exported and converted to the sequence spectrum list format using Excel (Microso ). Crosslink peak areas were assessed using the MS1 full-scan ltering protocol for peaks within 10 min of the crosslink spectral match identi cation. Peak areas were assigned to the speci ed crosslinked peptide identi cation if the mass error was within 10 ppm of the theoretical mass, if the isotope dot product was greater than 0.9, and if the peak was not found in the non-crosslinked negative control samples. e isotope dot product compares the distribution of the measured MS1 signals against the theoretical isotope abundance distribution calculated based on the peptide sequence. Its value ranges between 0 and 1, where 1 indicates a perfect match 81 . Pairwise comparisons were made using the "MSstats" package 82 implemented in Skyline to calculate relative fold changes and signi cance. Signi cant change thresholds were de ned as a log2 fold change ± 2 and -log10 p-value greater than 1.3 (i.e., a p-value less than 0.05). e visualization of proteins and crosslinks was generated . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint using xiNET 83 .

Fluorescent labeling of protein samples for smFRET
Puri ed PPARγ AF-1 D33C+Q121C mutant protein at 50 µM was labeled with Alexa Fluor 488 and Alexa Fluor 647 maleimide dyes ( ermo Fisher) in 20 mM sodium phosphate pH 7.2, 50 mM NaCl, 1 mM TCEP using a 5 mM dye stock. Six substoichiometric additions of the dyes were made to the protein construct over 3 hrs for a nal threefold molar excess of each dye vs protein. e labeled sample was passed twice through Zeba desalting columns ( ermo Scienti c) equilibrated in the same bu er to remove excess uorophores.

Single-molecule FRET (smFRET)
smFRET was performed using a two-color approach with a confocal setup where uorescently labeled proteins di use freely. To record smFRET data, the labeled protein at 15 µM was diluted 100,000-fold into the bu er (10 mM sodium phosphate with speci ed pH values and NaCl concentrations) reaching a nal concentration of approximately 150 pM. e red laser was delayed by ~20 ns with respect to the blue laser. Linear polarization was cleaned up (Glan-Taylor Polarizer, orlabs, GT10-A) and the red and blue light were combined into a single-mode optical ber (kineFlex, Point Source) before the light (100 µW of 483 nm light and 75 µW of 635 nm light) was re ected into the back port of the microscope (Axiuovert 200, Zeiss) and to the objective (C-APOCHROMAT, 40x/1.2 W, Zeiss). Sample emission was transmitted through a polychroic mirror (Chroma, ZT488/640rpc), focused through a 75-mm pinhole and spectrally split (Semrock FF593-Di03-25x36). e blue range was ltered (Semrock, FF03-525/50-25), and polarization was split (PBS101, orlabs) into parallel and perpendicular channels. e red range was also ltered (Semrock, FF01-698/70-25), and polarization was split (PBS101, orlabs). Photons were detected on four avalanche photodiodes (SPCM AQR 14, PerkinElmer Optoelectronics, for the green parallel and perpendicular channels and for the red parallel channel and SPCM AQRH 14, Excelitas, for the red perpendicular channel), which were connected to a time-correlated single-photon counting (TCSPC) device (MultiHarp 150N, PicoQuant). Signals were stored in 12-bit rst-in-rstout (FIFO) les. Microscope alignment was carried out using uorescence correlation spectroscopy (FCS) on freely di using ATTO 488-CA and ATTO 655-CA (ATTO-TEC). Instrument response functions (IRFs) were recorded one detector at-a-time in a solution of ATTO 488-CA or ATTO 655-CA in near-saturated centrifuged potassium iodide at a 25-kHz average count rate for a total of 25 × 106 photons. Macro Timedependent microtime shi ing was corrected for two (blue/parallel and red/perpendicular) of four avalanche photodiodes (APDs) using the IRF data as input. Data were analyzed using PIE Analysis with Matlab (PAM) so ware 84 using standard procedures for MFD-PIE smFRET burst analysis 85,86 . Signals from each TCSPC routing channel (corresponding to the individual detectors) were divided in time gates to discern 483-nm excited FRET photons from 635-nm excited acceptor photons. A twocolor MFD all-photon burst search algorithm using a 500-µs sliding time window (minimum of 100 photons per burst, minimum of 5 photons per time window) was used to identify single donor-and/or acceptor-labeled molecules in the uorescence traces. Double-labeled single molecules were selected from the raw burst data using a kernel density estimator (ALEX-2CDE ≤ 15) that also excluded other artifacts 87 . Sparse slowdi using aggregates were removed from the data by excluding bursts exhibiting a burst duration > 12 ms. By generating histograms of E versus measurement time, we corroborated that the distribution of E was invariant over the duration of the measurement. Data was corrected in this order to obtain the absolute stoichiometry parameter S and absolute FRET e ciency E: background subtraction, donor emission crosstalk correction, acceptor direct excitation correction and relative detection e ciency correction. To obtain the relative detection e ciency correction factor (γ), the center values of the E-S data cloud for each protein were estimated manually, plotted in an E vs. 1/S graph, and the data were t to the following equation where Ω is the intercept and Σ the slope of the linear t: γ= (Ω-1)/(Ω+Σ-1)

Luciferase reporter assays
HEK293T cells (ATCC #CRL-11268) were seeded in a white 96-well plate at 15,000 cells/mL per well. e following day, cells were transfected using Lipofectamine 2000 ( ermo Fisher Scienti c) and Opti-MEM with full-length PPARγ2 WT, ΔAF1, W39A, P40A, or empty vector control (pcDNA3.1) expression plasmids (5, 15 or 45 ng for each plasmid with 40, 30 or 0 ng of empty vector respectively to have the same amount of DNA per condition) along with tk-LUC 3X-PPRE luciferase reporter plasmid (45 ng) to a total of 90 ng per well and incubated for 18 hrs at 37 °C, 5% CO 2 . e media was aspirated without disturbing the cells then replaced with media supplemented with either 1 μM Rosiglitazone or the same volume of 100% DMSO and incubated 18 hrs at 37 °C, 5% CO 2 . e cells were harvested for luciferase activity quanti ed using Britelite Plus (Perkin Elmer; 25 μL) on a Synergy Neo plate reader (Biotek). Data were plotted as mean ± s.d. in GraphPad Prism; statistics performed using two-way ANOVA with Tukey multiple comparisons analysis and are representative of ≥2 independent experiments.

Western blot analysis
Six wells from each 45 ng HEK293T luciferase reporter assay transfection condition were washed with PBS, pooled, and harvested by centrifugation. Pellets were lysed in TN-T bu er (0.1 M Tris HCl, pH 7.5, 150 mM NaCl, 1% Triton X-100), centrifuged to remove the insoluble fraction, and the total protein in each sample was quanti ed using a Bradford assay. 40 μg protein/well was run on a 10% SDS-PAGE gel and transferred to PVDF membrane. Membranes were blocked in 5% milk and incubated overnight at 4 ˚C in primary antibody (anti-PPARγ, Santa Cruz #sc-7273; or anti-β-actin, Cell Signaling #3700). Washes were performed in TBS-T (20 mM Tris, 150 mM NaCl, 0.1% Tween-20). Secondary antibody (Jackson ImmunoResearch #115-005-003) was incubated for 1 hr at room temperature. Chemiluminescence was analyzed with SuperSignal West Femto Maximum Sensitivity Substrate ( ermoFisher #PI34095) for PPARγ or luminol reagent (Santa Cruz #sc-2048) for β-Actin using a Bio-Rad ChemiDoc Touch Imaging System. Figures were prepared using Bio-Rad Image Lab So ware.

Computational analyses
Disordered structural predictions were performed with ODiNPred 88 using the human PPARγ2 isoform (505 residues). AF-1 secondary structure propensity calculation was performed with SSP 29 using AF-1 Cα and Cβ NMR chemical shi assignments. e net charge per residue of the AF-1 sequence was calculated using localcider (http://pappulab. github.io/localCIDER/) 89 . Predicted PRE pro les from an extended AF-1 conformational ensemble (n=10,000 structures) with no long-range contacts were calculated using exible-meccano 32 .

ACKNOWLEDGMENTS
is work was supported by National Institutes of Health (NIH) grants . CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint R01DK124870 (to D.J.K.) from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK/DK); R01AG071332 (to P.R.G.) from the National Institute of Aging (NIA/AG); and R35GM130375 (to A.D.) from the National Institute of General Medical Sciences (NIGMS/ GM). e purchase of the 600 MHz NMR instrument was supported in part by NIH grant S10OD021550 from the NIH O ce of the Director. S.M. was supported by an NIH/NIDDK predoctoral fellowship (F31DK127643). K.-T.K. was supported by a Farris Foundation Fellowship. D.S. was supported by a postdoctoral fellowship from the Belgian American Educational Foundation. is content is solely the responsibility of the authors and does not necessarily represent the o cial views of the NIH.

ACCESSION NUMBERS
NMR chemical shi assignments of human PPARγ2 AF-1 (residues 1-135) have been deposited in the Biological Magnetic Resonance Data Bank (BMRB 51507). XL-MS data of full-length PPARγ and ΔAF-1 PPARγ have been deposited in Proteomics Identi cation Database (PRIDE PXD035304).
. CC-BY 4.0 International license available under a was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprint (which this version posted July 13, 2022. ; https://doi.org/10.1101/2022.07.13.499031 doi: bioRxiv preprint