Delineation of the molecular determinants of the unique allosteric binding site of the orphan nuclear receptor ROR g t

Nuclear receptors (NRs) are high-interest targets in drug discovery because of their involvement in numerous biological processes and diseases. Classically, NRs are targeted via their hydrophobic, orthosteric pocket. Although successful, this approach comes with challenges, including off-target effects due to lack of selectivity. Allosteric modulation of NR activity constitutes a promising pharmacological strategy. The retinoic acid receptor-related orphan receptor- g t (ROR g t) is a constitutively active NR that positively regulates the expression of interleukin-17 in T helper 17 cells. Inhibiting this process is an emerging strategy for managing autoimmune diseases. Recently, an allosteric binding pocket in the C-terminal region of the ligand-binding domain (LBD) of ROR g t was discovered that is amenable to small-molecule drug discovery. Compounds binding this pocket induce a reorientation of helix 12, thereby preventing coactivator recruitment. Therefore, inverse agonists binding this site with high affinity are actively being pursued. To elucidate the pocket formation mechanism, verify the uniqueness of this pocket, and substantiate the relevance of targeting this site, here we identified the key characteristics of the ROR g t allosteric region. We evaluated the effects of substitutions in the LBD on coactivator, orthosteric, and allosteric ligand binding. We found that two molecular elements unique to ROR g t, the length of helix 11 9 and a Gln-487 residue, are crucial The FA of the at very and the K at and , respectively, reveal their loss of coactivator affinity. Together, these data indicate that, except for the mutations in H12, the other constructs feature intrinsic coactivator affinities that allow the subsequent evaluation of inverse agonist binding. with the supplied protocol. The mutations were confirmed by sequencing using T7 promoter and T7 terminator primers (BaseClear).

The nuclear receptor (NR) superfamily in humans consists of 48 transcription factors that are essential for various physiological processes, and the dysregulation of these proteins occurs in many diseases (1).Within this family, there is a high level of homology between the members that all have a similar domain structure.The ligand-binding domain (LBD) of NRs integrates endogenous and exogenous signaling into a transcriptional output and binds to ligands via a three-layer antiparallel alpha-helical sandwich where the middle layer of helices is absent from the lower half of the domain, creating a cavity (2, 3).Because of the hydrophobic character of this cavity, most NRs interact with lipid-soluble ligands (3) that can alter the conformational equilibrium of this domain with concomitant modulation of cofactor recruitment and changes in target gene expression (4).While highly successful, targeting of this orthosteric binding site by drugs comes with challenges in selectivity, side effects, disease-related mutations, and resistance occurrence (5)(6)(7).In addition to developing orthosteric partial (ant)agonists (8,9), allosteric modulation of NR activity has been brought forward as a novel pharmacological strategy, potentially overcoming some of the challenges connected to orthosteric ligand targeting (10)(11)(12).Several examples of NR allosteric binding sites have been reported, in part overlapping the orthosteric binding site (13)(14)(15).
An intriguing allosteric site, highly promising for smallmolecule targeting, was identified for the retinoic acid receptor-related orphan receptor-gt (RORgt) (16).RORgt is a constitutively active orphan NR (17) that positively regulates the expression of interleukin-17 (IL-17) in Th17 cells (18)(19)(20).Direct inhibition of IL-17 in autoimmune diseases using monoclonal antibodies has already been proven to be effective in psoriasis, with three FDA-approved drugs in the clinic (21)(22)(23) and with other clinical applications further emerging (24).Smallmolecule inhibition of RORgt is similarly of high interest, and inhibitory inverse agonists, also of the orthosteric type, are intensely pursued (25,26).
Ligands that bind to the allosteric pocket in the C-terminal region of the LBD of RORgt also act as inverse agonists (16,(27)(28)(29)(30) and appear to do so with high NR isoform selectivity.They inhibit cofactor binding by inducing a reorientation of RORgt helix 12 (H12) incompatible with accommodating the classical LXXLL coactivator motif (Fig. S1).Upon allosteric ligand binding, the helix 11 prime (H119), the alpha-helical linker that connects helix 11 (H11) and H12 and a unique feature of the ROR family, unfolds and allows reorientation of H12.Together, this generates a cavity between helices 3 (H3), 4 (H4), H11, H119, and H12.The resulting RORgt conformation antagonizes cofactor binding in a manner different from "classical" orthogonal targeting, which results in destabilizing H12 folding (31).A variety of compounds targeting this allosteric pocket have recently been published (16,(27)(28)(29)(30) and effects thereof on autoimmune models studied (32).Notwithstanding this, the mechanism underlying the formation of the allosteric pocket in RORgt, the prerequisites thereof, and the NR isoform specificity are poorly understood.Delineation of the molecular determinants of this RORgt allosteric binding site is crucial for addressing these issues and for underlining the developmental potential of targeting this site.
Here, we performed structure and sequence alignments to identify unique elements and key differences between RORgt and other NRs in the allosteric region.Crucial amino acids within the allosteric site and those connecting distant helices in the allosteric fold were evaluated.The roles of these specific RORgt elements were subsequently studied in the context of their response to both orthosteric and allosteric ligands to obtain a comprehensive understanding of their role in the formation of the allosteric pocket and implications for drug development and NR selectivity.

Identification of unique RORgt elements aligned with allosteric pocket formation
An in-depth structure analysis and multiple-sequence alignments were executed to identify amino acids crucial for RORgt allosteric pocket formation.Residues within 4 Å of the allosteric ligand MRL-871 in the RORgt crystal structure (PDB entry 4YPQ) (16) were examined and considered for mutational studies (Fig. 1 boldface letters).Residues pointing out of the allosteric pocket and not engaged in obvious interactions between secondary structure elements were excluded (Leu-324, Val-480, Leu-483, Phe-498, and Tyr-502).The Dali server (33) was used for structural alignment of the RORgt LBD (PDB entry 3KYT) (34) against the full PDB (33,34).The list of resulting structures was checked for human NRs, and all 39 of the other human NRs for which an LBD crystal structure is available could be identified (for RORb, the structure of the 98% similar protein derived from Rattus norvegicus was used).For each NR, the structure with the highest similarity score compared with the RORgt structure was chosen, and a multiple-sequence alignment based on the structural information was made.The eight NRs for which no structural information about their LBDs was available were added to the alignment afterward using the MAFFT online service (35), which was also used to create a phylogenetic tree based on the sequence similarities (Fig. 2) (35,36).The alignment was used to determine the level of conservation of the amino acids in the allosteric pocket (Fig. S2).Amino acids of RORgt that were over 30% conserved within the NR superfamily (Thr-325, Leu-353, and Lys-354) and Ile-328 and Leu-505, which have a similarly sized hydrophobic residue at this position in most NRs, were not mutated.Alanines in the allosteric pocket were also not considered for mutations (Ala-321, Ala-496, and Ala-497).As a result, residues Trp-317, Gln-329, Gln-484, Gln-487, Glu-504, and Phe-506, as well as the unique H119 Val-494-Gln-495 stretch, were identified for in-detail analysis (Fig. 1).
The tryptophan at position 317 in RORgt makes a hydrogen bond to His-490 in H119 in the allosteric fold (Fig. 1A).This tryptophan is uniquely present in the ROR, retinoic acid receptor (RAR), and Rev-erb family, while 77% of NRs do not have a hydrogen bond donor at this position.Since the most abundant residue at this position is a phenylalanine, a W317F construct was designed to elucidate the role of the hydrogen bond in forming the allosteric pocket.Gln-329 appears to fulfill a dual role within the allosteric pocket by connecting H3 and H119 and by making a hydrogen bond to the carboxylic acid of the allosteric ligand (Fig. 1B).Deleting these hydrogen bonds was hypothesized to correspond to a loss of affinity for allosteric ligands.To test this, and because this position has highly variable amino acids in other NRs, a Q329A mutation was introduced.Gln-484 makes a hydrogen bond to H12 in both the agonistic (Thr-508) and allosteric (Leu-505) fold (Fig. 1C), potentially aiding in the stability of both folds.Interestingly, this amino acid is not conserved within the ROR family.To elucidate its role in allosteric pocket formation, a Q484A mutation was made.Gln-487 forms two hydrogen bonds to the backbone around Ile-492 in the unfolded H119, which seems to help to correctly position this loop over the allosteric pocket (Fig. 1D).The same glutamine residue is only present in the Pregnane X receptor (PXR), while 50% of NRs, including RORa and RORb, have a lysine at this position that would not be able to make both hydrogen bonds.To evaluate the specific role of this glutamine compared with that of the other NRs, a Q487K mutation was made.Phe-506 makes a p-p stacking interaction with the allosteric ligand (37), which might be a prerequisite for binding (Fig. 1H); therefore, an F506A mutation was made.A final feature, which might be key in forming the allosteric pocket, is the length of H119.In the ROR family, there are 10 residues between H11 and H12, while in all other NRs, this region has a maximum length of 8 residues.The crucial nature of the length of H119 for correctly positioning H12 was investigated using two RORgt constructs featuring a single (Q495del) or double (V494-Q495del) deletion (Fig. 1, E-F).The deleted residues were selected in such a way that no specific side-chain interactions would be lost.A final control construct was made that should be able to bind to compounds in the same way as the WT protein but without the ability to bind to a coactivator sequence.The E504A mutation removes one side of the charge clamp (38), necessary for coactivator binding (Fig. 1G).

Coactivator affinity is only affected by helix 12 mutations
A fluorescence anisotropy (FA) assay was employed to determine the unliganded coactivator affinity of the different RORgt constructs as a means to establish correct folding and basal level toward ligand response.The WT RORgt LBD showed an affinity to the FITC-labeled SRC1b2 coactivator peptide of 14.5 6 4.53 mM (Table 1 and Fig. S3).The Q487K and Q495del constructs demonstrated coactivator affinities in the same range, with K d values of 9.65 6 0.95 mM and 12.2 6 5.3 mM, respectively.The Q484A construct bound somewhat more strongly to the coactivator peptide than the WT protein, with a K d value of 4.87 6 3.69 mM, indicating a high basal affinity.Q329A and V494-Q495del showed a slightly decreased intrinsic affinity compared with that of the WT protein, with K d values of 34.3 6 11.7 mM and 33.7 6 9.1 mM, respectively.The RORgt W317F construct  (33,35,36).On the top, a 2D outline of the RORgt LBD is given with helices numbered according to convention.Boldface stripes and letters indicate regions within 4 Å of the allosteric compound (PDB entry 4YPQ).Cursive stripes and letters indicate regions within 4 Å of the orthosteric compound (PDB entry 3KYT) (note the overlap between both binding sites in H3, H4-5, and H12).Crystal structures that were used for the structural alignment are indicated with their 4-digit PDB code.NRs for which no LBD crystal structures were available are indicated.Residues chosen for mutagenesis are highlighted in different colors, and their connected colored symbol for easy identification in subsequent studies is listed.
bound even weaker and had a K d value of only 78.6 6 10.4 mM.The FA signal of the E504A and F506A constructs only increased at very high protein concentrations, and the K d values at 200 6 36 mM and 365 6 292 mM, respectively, reveal their loss of coactivator affinity.Together, these data indicate that, except for the mutations in H12, the other constructs feature intrinsic coactivator affinities that allow the subsequent evaluation of inverse agonist binding.

Ligand binding in the RORgt orthosteric pocket is not affected by mutations in the RORgt allosteric pocket
The functionality of the different RORgt LBDs in response to ligands was evaluated using a known orthosteric agonist and inverse agonist.For this, 20a-hydroxycholesterol and ursolic acid (UA) were chosen (Fig. S1), as their binding modes are well known and constricted to the classical, orthosteric binding pocket (PDB entry 5K3M) (34).A homogeneous time-resolved FRET (HTRF) assay and differential scanning fluorimetry (DSF) were utilized for this.In HTRF, the RORgt LBDs were labeled using a terbium-labeled His 6 -tag binding antibody that functions as a fluorescent donor.Biotinylated SRC1b2 coactivator peptide with streptavidin D2 was used as the fluorescent acceptor.Upon increasing the 20a-hydroxycholesterol concentration, an increase in FRET ratio is anticipated, while the addition of UA will decrease this ratio.In Fig. 3 (left and middle), it is shown that this holds true for all constructs except the H12 E504A and F506A proteins, which have no affinity for the coactivator peptide.EC 50 or IC 50 values were found in the same range for the WT and mutated RORgt LBDs for these compounds (Table S1).The difference in maximum HTRF ratio increase that is obtained for the V494-Q495 construct by 20ahydroxycholesterol might be related to a different positioning of H12 by the shorter H119.The differences in HTRF ratio decrease caused by UA are caused by differences in starting levels without ligand (Fig. S4).All HTRF ratios go down to the level that is also reached in the absence of protein.
To orthogonally validate the orthosteric ligand binding effects in a coactivator-independent manner, DSF studies were also performed with these compounds on all the protein constructs (Fig. 4, Fig. S5, and Table S2).All proteins are stabilized both by binding to the orthosteric inverse agonist UA, by an average of 6.7 6 0.6 °C, and by binding the orthosteric agonist 20a-hydroxycholesterol (4.3 6 0.8 °C).The data also clearly allow concluding that even though the E504A and F506A con-structs are not able to efficiently bind to a coactivator, they are still capable of binding to the orthosteric ligands, as they show strong stabilization upon addition of these compounds.Together, these findings show that all protein constructs have a functioning orthosteric pocket that is not affected by the mutations around the allosteric pocket.
Helix 119 and Q487 are crucial for formation of the RORgt allosteric pocket and allosteric ligand response The response of the different RORgt to an allosteric inverse agonist was also evaluated using HTRF and DSF assays.In HTRF, the allosteric inverse agonist MRL-871 inhibited the coactivator binding of WT RORgt with an IC 50 of 4.90 6 1.98 nM (Fig. 3, right).Three RORgt constructs matched this nanomolar inhibitory activity, being W317F with an IC 50 of 3.09 6 1.97 nM and Q329A and Q484A with IC 50 values of 13.1 6 4.6 nM and 28.5 6 11.5 nM, respectively.RORgt Q487K interaction with the allosteric ligand was over 500 times weaker than that with the WT RORgt LBD, with an IC 50 of 2.75 6 0.65 mM.The two deletion constructs hardly showed binding to the allosteric compound, with IC 50 values of only ;50-100 mM.Again, the difference in bottom plateaus is caused by the difference in starting levels because of the varying coactivator affinity (Fig. S4).
The findings regarding coactivator recruitment were reflected in the DSF studies (Fig. 4, Fig. S5, and Table S2).The WT RORgt LBD and E504A construct, which could not be measured in HTRF due to its lack of coactivator binding, both showed an increase in melting temperature (T m ) of 7.7 6 0.2 °C in the presence of the allosteric inverse agonist.The W317F, Q329A, Q484A, and F506A constructs were also stabilized upon allosteric ligand binding, albeit somewhat weaker, by 5.8 6 0.5 °C, 5.1 6 0.2 °C, 5.8 6 0.3 °C, and 3.4 6 0.5 °C, respectively.Interestingly, Q487K had the exact same melting temperature in the absence and presence of MRL-871, meaning that this protein was not stabilized upon the addition of the allosteric ligand.The deletion constructs even showed a decrease in the T m upon MRL-871 treatment of 21.6 6 0.1 °C for both the single and double deletion.When plotting the DT m s determined in the DSF assay against the IC 50 values from the HTRF (Fig. S6), a strong correlation can be observed, pointing to correlated mechanisms of allosteric pocket modulation observed via the two different assay formats.

Discussion
The most N-terminal amino acid evaluated (Trp-317) reveals an important contribution to protein activity and stability.The W317F mutation leads to a lower intrinsic coactivator affinity of RORgt (Table 1, Fig. S3) and lowers the protein stability (Fig. S7) compared to that of the WT protein.Crystal structure analysis reveals that in the apo (39) and agonist-bound (34) state, Trp-317 contributes to a hydrophobic network, as extensively investigated by Sun et al. (40), and the indole nitrogen also serves as a hydrogen bond donor toward the backbone amide of L391 in H7-8.These characteristics illustrate why the W317F protein is less stable and active.The allosteric inverse agonist has effects on W317F similar to those of the WT protein in terms of inverse agonism and conferring protein stability (Fig. 3-4).Together, this shows that the hydrogen bond between Trp-317 and His-490, and Trp-317 in general, are not of relevance for the formation of the allosteric pocket.
The construct with the Q329A mutation binds slightly less strongly to the FITC-labeled SRC1b2 peptide and is about 1 °C less thermostable than the WT RORgt (Table 1, Fig. S7).In the HTRF, the IC 50 value for the allosteric inverse agonist was in the same range as that for the WT protein (Fig. 3), and the DSF results also report a strong thermal stabilization of the construct upon addition of the allosteric ligand (Fig. 4).Yuan et al. conducted molecular dynamics simulations on RORgt bound to MRL-871 and reported that the hydrogen-bonding interaction between Gln-329 and the ligand, as observed in the crystal structure, is only present during 1.5% of their simulation time (41).Thus, these simulations corroborate the experimental findings from this study that Gln-329 is not crucial for binding of the MRL-871 allosteric ligand.Analogously, Gln-329 can be considered not to be a crucial determinant for forming the RORgt allosteric pocket.
The mutation of the RORgt Gln-484 to an alanine slightly increases the affinity of the LBD toward the coactivator peptide.In the crystal structure of the agonist-liganded and apo RORgt, the Gln-484 sidechain appears to make a hydrogen bond to Thr-508 in H12, but closer inspection of the actual electron densities reveals that the orientation of this sidechain is unclear.In contrast, in the crystal structure of RORgt LBD bound to the allosteric inverse agonist, the electron density of the sidechain of Gln-484, which makes a hydrogen bond with Ser-507, is highly defined.The Q484A construct is indeed somewhat less susceptible to allosteric inverse agonist binding than the WT protein, with the IC 50 value in the HTRF assay being ;6 times weaker.A similar effect is seen upon titration of allosteric compound FM26 (Fig. S8) (30).Concomitantly, MRL-871 still induces strong thermostability of the protein but to a lesser extent than that observed for the WT protein.Thus, these results highlight that, while not crucial, Gln-484 contributes to the formation of the allosteric pocket, most probably via hydrogen bonding to H12.
Gln-487 revealed itself to be highly crucial for the formation of the allosteric pocket.In the assays designed to validate the classical orthosteric regulation of the RORgt LBD, the Q487K construct showed behavior very similar to that of the WT protein.The thermal stability of the apo Q487K construct is even a bit higher than that of the WT, and orthosteric ligand binding, both of the agonist and inverse agonist types, leads to similar coactivator recruitment and inhibition effects.These observations are in line with the design expectations, since the singular hydrogen bond that Gln-487 makes with Ser-508 in H12 in the agonistic fold can most probably be made with the lysine mutation as well.In striking contrast to the absence of effects on the orthosteric pocket, the Q487K construct featured very pronounced differences from the WT RORgt regarding its allosteric pocket.In the HTRF assay, the IC 50 value for MRL-871 inhibition of the Q487K construct is weakened by a factor of 560, and the shift for allosteric compound FM26 appears to be even bigger (Fig. S8) (30).In the DSF assay, no thermal stabilization at all is observed upon dosing of the allosteric ligand.The glutamine at the 487 position is highly unique within the NR superfamily and only shared with PXR, which is otherwise only 28% identical.Half of all NRs, including RORa and RORb, have a lysine at this position.These results and observations are highly relevant, as this implies that, on the one hand, the allosteric pocket formation is uniquely attributable to RORgt via this glutamine and that, on the other hand, selective ligand targeting of RORgt is ideally suited via this allosteric pocket.
The two RORgt constructs with a single or double amino acid deletion in the H119 allowed the evaluation of the importance of this longer linker between H11 and H12, unique to the ROR family.The results from the FA assay showed that the shortening of the H119 did not affect the capacity of RORgt to bind to a coactivator peptide (Table 1, Fig. S3).Furthermore, the orthosteric ligands, both of agonistic and inverse agonistic nature, induced coactivator binding and thermal stability effects in line with those for the WT protein.In outspoken contrast to the above, the addition of the allosteric inverse agonist showed a completely different profile on the two H119 deletion constructs from that of the WT protein.In the HTRF assay, MRL-871 only inhibited both proteins at high micromolar concentrations, and in DSF the allosteric ligand even destabilizes these proteins.The crystal structure of RORgt bound to MRL-871 reveals that with the formation of the allosteric pocket, H119 extends to form a loop that allows H12 to be positioned overlapping the site where coactivators can normally bind (Fig. S1).The shortening of this loop leads to constructs that cannot span this distance anymore, thereby preventing H12 from forming the allosteric pocket.This finding leads us to conclude that NRs that do not have this H119 might not be able to form a similar allosteric pocket.
The E504A construct was designed as a control protein with both the orthosteric and allosteric pocket functioning but without the ability to efficiently bind coactivator protein or peptide (38).The experimental results confirm this behavior, with the FA data reporting coactivator binding only at very high protein concentrations (Table 1, Fig. S3), while the DSF data report thermal stability and ligand response as a virtual copy of the WT protein (Fig. 4).
In the conformation the RORgt LBD adopts upon allosteric inverse agonist binding, Phe-506 is pointing directly into the allosteric pocket and is involved in a p-p stacking interaction with the MRL-871 ligand (16,41).In the apo or agonist-bound state, Phe-506 is involved in aromatic interactions with His-479 and Glu-502, providing an anchoring force to stabilize H12 in the active conformation even in the absence of ligand (39).The F506A construct indeed shows diminished coactivator peptide binding to an extent similar to that of the E504A construct (Table 1, Fig. S3), illustrating the importance of the aromatic cluster in the active, agonistic conformation.The orthosteric ligands do still bind just as well as to the WT protein, as evidenced by the thermal stabilization data (Fig. 4).The absence of coactivator binding makes the evaluation of MRL-871 in related assays impossible.The DSF assay reveals that the F506A construct still can bind and be stabilized by the allosteric inverse agonist, albeit to a lesser extent than the WT protein.This implies that the p-p stacking interaction between Phe-506 and MRL-871 is relevant but not essential for allosteric compound binding.A similar observation was recently made upon structural comparison of different RORgt allosteric ligands (37).Other NR superfamily members contain various medium-sized hydrophobic or aromatic residues at this position.Since even the F506A mutation retains affinity toward the allosteric compound, this specific amino acid is deemed not crucial for the formation of the allosteric pocket but potentially modulatory toward allosteric ligand binding.
In conclusion, the molecular determinants of the RORgt allosteric pocket were evaluated in detail by selectively introduced mutations to the RORgt LBD.All RORgt proteins were still responsive to orthosteric agonists and inverse agonists, demonstrating their overall functionality.From the set of proteins studied, two molecular elements were crucial for the formation of the allosteric pocket of RORgt.The combination of Gln-487 and a long H119 are unique within the NR realm for RORgt.Whereas other amino acids that delineate the allosteric pocket can be replaced for those with other side chains, modifications to these two elements are detrimental.Thus, the results of this study strongly bring forward that the allosteric pocket, as present in RORgt and receptive to MRL-871 and other ligands (16,(27)(28)(29)(30)37), is highly unique to RORgt within the NR superfamily.This implies that RORgt is the only NR capable of forming this specific type of allosteric pocket, which is also supported by a commercially available panel of cell-based NR reporter assays against which an MRL-871 analog was tested (16).This uniqueness testifies to the potential of achieving high NR subtype selectivity via drugging of this allosteric RORgt pocket for applications in Th17-mediated autoimmune diseases.

Site-directed mutagenesis
Point mutations (W317F, Q329A, Q484A, Q487K, E504A, and F506A) were introduced using the QuikChange Lightning multisite-directed mutagenesis kit (Agilent) in accordance with the protocols described in the kit manual.The parental DNA for this was a pET15b vector containing the RORgt LBD (residues 265-518) with an N-terminal His 6 tag.Single primers designed to hybridize with the parental DNA containing 1-3 mismatches were utilized for this (Table S3).
The primers for the deletion constructs, Q495del and V494_Q495del, were prepared via the method described by Liu and Naismith (42).In short, primers with overlap on the 59 end (pp) were designed in such a way that the deletion was in the pp region while the 39 ends (no) could hybridize to the DNA efficiently (Table S3).For the PCR reaction, 0.2 ng/ml parental DNA was combined with 1 mM both primers and 13 Phusion polymerase in HF buffer (Thermo Scientific).The reaction started with denaturation (1 min, 98 °C), followed by 18 amplification steps.Each of these steps started with denaturation (10 s, 98 °C), followed by annealing (1 min, T m no of 25 °C) and extension (2.5 min, 72 °C).The method ended with a final annealing (1 min, T m pp of 25 °C) and extension (30 min, 72 °C) step.
For all mutagenesis reactions, the parental DNA was digested using DpnI followed by transformation into XL10-Gold ultracompetent cells (Agilent) by heat shock.Small cultures were initiated using single colonies in 5 ml LB medium supplemented with ampicillin (100 mg/ml) overnight at 37 °C 250 rpm.DNA was isolated using the QIAprep spin miniprep kit (Qiagen) in accordance with the supplied protocol.The mutations were confirmed by sequencing using T7 promoter and T7 terminator primers (BaseClear).
RORgt-LBD SRC-1 coactivator protein titration FA assay FA assays were performed in 10-ml volumes in triplicate in buffer containing 150 mM NaCl, 10 mM HEPES, 5 mM DTT, 10 mM CHAPS, and 0.1%, m/v, BSA. 10 nM FITC-labeled SRC1b2 (Table S4) was present, and His 6 -RORgt LBD was titrated to the peptide in round-bottom, nonbinding, black, 384-well plates (Corning, 4511).The plates were measured in an Infinite F500 plate reader (Tecan) (excitation, 485 nm; emission, 535 nm).The data were then analyzed using GraphPad Prism software, where the curve-fitting was done via y 5 Where B max is the maximum specific binding, constrained to a maximum value of 350, K d is the equilibrium binding constant, and background is the amount of nonspecific binding with no added protein.Background was assumed to be constant between constructs within each experiment.The reported K d values are the means 6 S.D. from five individually fitted independent experiments performed singly.

RORgt-LBD SRC-1 coactivator compound titration HTRF assay
HTRF assays were performed in 10-ml volumes in triplicate in buffer containing 150 mM NaCl, 10 mM HEPES, 5 mM DTT, 10 mM CHAPS, and 0.1%, m/v, BSA.20 nM His 6 -RORgt LBD, 100 nM biotin-labeled SRC1b2 coactivator peptide (Table S4), 733 pM terbium-labeled anti-His antibody (Cisbio, 61HISTLB), and 12.5 nM D2-labeled streptavidin (Cisbio, 610SADLB) were also used.Compounds dissolved in DMSO were titrated to this mixture in round-bottom, nonbinding, white, 384-well plates (Corning, 4513), keeping the DMSO concentration constant at 2%.After 30 min of incubation at 4 °C, the plates were measured in an Infinite F500 plate reader (Tecan) (excitation, 340 nm; emission, 665 nm and 620 nm).The data were then analyzed using GraphPad Prism software, where the curve-fitting was done via y ¼ Bottom 1 ðTop À BottomÞ Top and Bottom are defined as the plateaus in the units of the Y axis, reflecting maximum and minimum binding levels.The hillslope was assumed to be constant within samples with the same compound to aid in the fit of datasets where no plateau was reached.For the measurements with 20a-hydroxycholesterol, data points at higher compound concentrations than that of the top, where the decrease in signal compared with the top was more than 10%, were omitted for the fit.The reported EC 50 /IC 50 values are the means 6 S.D. from three individually fitted independent experiments performed in duplicate.

DSF
Differential scanning fluorimetry (DSF) assays were performed using 40-ml samples containing 5 mM RORgt-LBD, 10 mM compound, and 2.53 SYPRO orange (Sigma) in buffer containing 150 mM NaCl, 10 mM HEPES, and 2% DMSO.The samples were heated from 30 °C to 64.5 °C at a rate of 0.3 °C per 15 s in a CFX96 touch real-time PCR detection system (Bio-Rad).Excitation (575/30 nm) and emission (630/40 nm) filters were used, and the reported melting values were calculated as the local minimum between 35 °C and 60 °C in the negative derivative of the resulting melting curve.T m and DT m are determined as means 6 S.D. from 3 independent experiments performed in duplicate.

Figure 2 .
Figure 2. Multiple-sequence alignment of the human NR LBDs based on structural alignment.On the left a phylogenetic tree is shown, based on the sequence alignment(33,35,36).On the top, a 2D outline of the RORgt LBD is given with helices numbered according to convention.Boldface stripes and letters indicate regions within 4 Å of the allosteric compound (PDB entry 4YPQ).Cursive stripes and letters indicate regions within 4 Å of the orthosteric compound (PDB entry 3KYT) (note the overlap between both binding sites in H3, H4-5, and H12).Crystal structures that were used for the structural alignment are indicated with their 4-digit PDB code.NRs for which no LBD crystal structures were available are indicated.Residues chosen for mutagenesis are highlighted in different colors, and their connected colored symbol for easy identification in subsequent studies is listed.

Figure 3 .
Figure 3. Representative HTRF binding curves of the WT and mutated proteins.Dose-response curves from the HTRF coactivator recruitment assay at fixed protein (20 nM) and SRC1b2 (100 nM) concentrations are shown.Left, 20a-hydroxycholesterol titration; middle, ursolic acid titration; right, MRL-871 titration.Bottom, schematic representation of the assay.Mean 6 S.D. from one experiment performed in duplicate is shown with the corresponding fit.

Figure 4 .
Figure 4. Average DT m values of ligand effects on the RORgt protein constructs in DSF.T m of protein (5 mM) is determined in the absence and presence of different ligands (10 mM).The T m without ligand is subtracted from that with ligand, yielding the DT m .Top left, schematic of the assay; graphs, data per protein.Graphs show the average DT m per experiment and mean 6 S.D. from 3 experiments performed in duplicate.

Table 1
K d values of the wild-type and mutated proteins toward FITC-labeled SRC1b2 peptide a a Values shown are means 6 S.D. from five independently executed and fitted experiments.