Structural and mechanistic insights into the interaction of the circadian transcription factor BMAL1 with the KIX domain of the CREB-binding protein

The mammalian CLOCK:BMAL1 transcription factor complex and its coactivators CREB-binding protein (CBP)/p300 and mixed-lineage leukemia 1 (MLL1) critically regulate circadian transcription and chromatin modification. Circadian oscillations are regulated by interactions of BMAL1's C-terminal transactivation domain (TAD) with the KIX domain of CBP/p300 (activating) and with the clock protein CRY1 (repressing) as well as by the BMAL1 G-region preceding the TAD. Circadian acetylation of Lys537 within the G-region enhances repressive BMAL1-TAD-CRY1 interactions. Here, we characterized the interaction of the CBP-KIX domain with BMAL1 proteins, including the BMAL1-TAD, parts of the G-region, and Lys537. Tethering the small compound 1–10 in the MLL-binding pocket of the CBP-KIX domain weakened BMAL1 binding, and MLL1-bound KIX did not form a ternary complex with BMAL1, indicating that the MLL-binding pocket is important for KIX-BMAL1 interactions. Small-angle X-ray scattering (SAXS) models of BMAL1 and BMAL1:KIX complexes revealed that the N-terminal BMAL1 G-region including Lys537 forms elongated extensions emerging from the bulkier BMAL1-TAD:KIX core complex. Fitting high-resolution KIX domain structures into the SAXS-derived envelopes suggested that the G-region emerges near the MLL-binding pocket, further supporting a role of this pocket in BMAL1 binding. Additionally, mutations in the second CREB-pKID/c-Myb–binding pocket of the KIX domain moderately impacted BMAL1 binding. The BMAL1(K537Q) mutation mimicking Lys537 acetylation, however, did not affect the KIX-binding affinity, in contrast to its enhancing effect on CRY1 binding. Our results significantly advance the mechanistic understanding of the protein interaction networks controlling CLOCK:BMAL1- and CBP-dependent gene regulation in the mammalian circadian clock.

to target both the MLL and the pKID/c-Myb pockets in a cooperative manner.
The amount of CBP protein present in vivo is limiting. Therefore, binding competition of different transcriptional regulators is crucial to define and regulate the diverse biological functions of CBP (17). Interestingly, the KIX ligands MLL1 and CREB also play important roles in the regulation of circadian transcription (31-33). MLL1 co-activates circadian genes by daily rhythmic recruitment to CLOCK:BMAL1 and cyclic H3 Lys 4 trimethylation, and this MLL1 activity is further enhanced by p300/CBP-dependent MLL1 acetylation (32,33). CREB-CBP interactions, on the other hand, are essential for light entrainment of the circadian master clock in the suprachiasmatic nucleus via up-regulation of the per1 clock gene, which contains CLOCK:BMAL1-and CREB-binding sites in its promoter (34,35). These observations suggest a functional interplay of CBP, BMAL1, MLL1, and CREB within circadian gene regulation, chromatin modification, and light entrainment, which, at a mechanistic level, is far from understood. Whereas the KIX-and CRY-binding BMAL1-TAD regions have been described (9,14), the binding sites of the BMAL1-TAD within the KIX domain and the interplay of BMAL1 with other KIX ligands are largely undefined. Here, we purified three C-terminal BMAL1 constructs including the BMAL1-TAD and parts of the G-region to unravel the mechanisms underlying their interaction with the CBP-KIX domain. We show that tethering the KIX domain with the small molecule 1-10 in the MLL-binding pocket lowers its affinity for BMAL1, suggesting a role of this pocket in BMAL1 binding. In a pulldown experiment, MLL1 sequestered KIX from a preformed KIX:BMAL1 complex, and therefore a ternary MLL1-TAD:BMAL1-TAD:CBP-KIX complex could not be formed. Moreover, mutations within the c-Myb/pKID-binding pocket of KIX decreased the affinity for BMAL1, suggesting that both KIX-binding pockets are involved in BMAL1 interactions. Furthermore, we showed that the BMAL1(K537Q) mutant mimicking lysine acetylation does not alter the BMAL1 affinity for KIX, contrary to its enhancing effect on the repressive CRY1-BMAL1 interaction (13,14). Our CD spectra suggest that BMAL1 does not become structured upon complexation with KIX, which contrasts with the binding-induced folding reported for other KIX ligands (19,20,23). Finally, we present small-angle X-ray scattering (SAXS) models of BMAL1 and three BMAL1:KIX complexes, which depict the BMAL1 regions N-terminal to the TAD including Lys 537 and parts of the G-region emerging from one side of the BMAL1-TAD:KIX core complex, thereby suggesting a directionality for the BMAL1-TAD-CBP-KIX interactions.
As described previously (16) and shown by our nano-surface plasmon resonance (NanoSPR) and microscale thermophoresis (MST) measurements (Fig. 4), the P624A mutation, which stabilizes the trans conformation of the Trp 623 -Pro 624 imide bond of BMAL1, does not affect the BMAL1 affinity to KIX. Due to its enhanced stability after purification, we therefore used the BMAL530(P624A) mutant protein for biophysical interaction studies with KIX.

BMAL1-CBP interactions in the circadian clock
about 5 M affinity. Notably, this affinity is somewhat lower than reported previously (16), possibly due to interference of the N-terminal BMAL1-His 6 tag or the RED-tris-NTA NT-647 fluorescent dye. Compared with WT KIX, the affinity of the untethered KIX(N627C) mutant protein to BMAL1 was 1.5-fold higher in MST and 3-fold higher in FP, whereas the L664C mutation showed a 1.5-fold reduced affinity in FP (Fig. 2). Tethering compound 1-10 to Cys 664 or Cys 627 reduced the BMAL1-binding affinity by about 2-2.5-fold compared with WT KIX in both our MST (Fig. 2, A, B, and E) and FP (Fig. 2, C-E) measurements, resulting in a BMAL1 affinity of about 10 M to both 1-10-KIX(L664C) and 1-10-KIX(N627C). These results suggest that the MLL-binding pocket plays a role in the interaction of KIX with BMAL1.

MLL1 displaces BMAL1 from a BMAL1:KIX complex
We purified a GST-tagged MLL1-TAD peptide (denoted as MLL19) of the MLL1 transactivation domain (TAD) that was used previously for quantitative biophysical and structural analyses of KIX-MLL1-TAD interactions (25,26). We performed a GST pulldown assay with GST-MLL19 as bait using either the purified preformed BMAL530:KIX complex as prey or simultaneously adding the two individually purified BMAL530 and KIX proteins as prey (Fig. 2F). Interestingly, MLL19 only pulled down KIX, but not BMAL1, when incubated with both KIX and BMAL1 proteins (Fig. 2F, lane 3), and MLL19 even sequestered KIX when incubated with the preformed BMAL1:KIX complex (Fig. 2F, lane 5). Notably, this . The normalized fluorescence (MST) or polarization (FP) values were plotted against the ligand concentration, and best fits were obtained using a one-site binding model. A ϳ2-fold lower BMAL1 affinity of 1-10-tethered KIX is observed with both methods, suggesting a role of the MLL-binding pocket in BMAL1 interactions. E, dissociation constants (K D ) for BMAL1-KIX interactions obtained from MST and FP. The values are the mean of two or three independent measurements. F, GST pulldown assay of KIX and BMAL530 (prey) with the MLL1-TAD peptide (GST-MLL19, bait). The inability of MLL1-bound KIX to retain BMAL1 in a ternary MLL1:KIX:BMAL1 complex confirms the importance of the MLL-binding pocket in BMAL1:KIX complex formation. Error bars, S.E.

BMAL1-CBP interactions in the circadian clock
effect was not observed in control pulldowns using GST only (Fig. 2F, lanes 2 and 4). Hence, a ternary complex between the BMAL1-TAD, the MLL1-TAD, and the CBP-KIX domain could not be formed under our pulldown conditions, which further supports the importance of the MLL-binding pocket for BMAL1 interactions.

Mutations in the MLL-and c-Myb/CREB-pKID-binding pocket of KIX affect BMAL1 binding
To further confirm the role of the MLL-binding pocket and to characterize the impact of the second c-Myb/CREB-pKIDbinding pocket on KIX-BMAL1 interactions, we also determined the effect of compound 1-10 and of mutations in both KIX domain pockets on BMAL1-binding affinities using single plasmonic nanoparticles as nanosensors, a technique termed NanoSPR, in analogy to the established and related surface plasmon resonance (SPR) method (36 -38). The NanoSPR technique was used for the comparative analysis of both KIX binding pockets, as it allowed us to measure the effects of 1-10 tethering and of all of our mutations in the MLL-and c-Myb/ pKID-binding pockets simultaneously within one titration experiment to make them internally comparable. In this Nano-SPR experiment, nanoparticles functionalized with N-terminally His-tagged WT or mutant KIX proteins were utilized as sensors. Increasing concentrations of the untagged BMAL530(P624A) protein were passed through the flow cell until the interaction reached equilibrium. Thereby wavelength shifts of about 1400 nanoparticles were measured to obtain 80 -200 data points (particles) (i.e. high statistics) for each WT or mutant KIX protein. Collectively, NanoSPR enabled us to also measure moderate (e.g. 2-fold) BMAL1 affinity changes caused by alterations in the two KIX-binding pockets within the same experiment with high statistics and using relatively low amounts of unlabeled proteins. In our NanoSPR experiment, the WT KIX domain was found to bind BMAL1 with an affinity of ϳ0.6 M, which is in the range determined in a previous study (16), likely because BMAL1 was not N-terminally Histagged or labeled. The L664C mutation in the MLL-binding pocket weakened the BMAL1 affinity by about 1.5-fold, whereas the N627C mutation showed nearly WT behavior (Fig. 3, A, B, and D). Tethering compound 1-10 to either Cys 664 or Cys 627 reduced the BMAL1 affinity by about 2-fold compared with WT KIX, resulting in affinities of ϳ1.5 and ϳ1.3 M for 1-10-KIX(L664C) and 1-10-KIX(N627C), respectively. These results are in qualitative agreement with our MST and FP measurements and provide further evidence that the MLL-binding pocket plays a role in the KIX-BMAL1 interaction. Notably, compound 1-10 weakens the binding of MLL1 to the KIX domain much more significantly, reducing MLL1-TAD affinity by about 100-fold (27), suggesting that BMAL1 uses a binding mode different from that used by MLL1. Furthermore, BMAL1 binding may be stabilized by additional binding interfaces (e.g. in the second KIX-binding pocket).
To analyze the role of the second c-Myb/pKID-binding pocket in BMAL1 interactions, we used NanoSPR to determine BMAL1-binding affinities to KIX proteins with a single Y658A and a double Y650A/A654W mutation in the shallow groove of this pocket (Figs. 1C and 3C). The mutations were chosen based

BMAL1-CBP interactions in the circadian clock
on their ability to significantly reduce the binding affinity of the KIX domain for c-Myb and pKID (21). Both mutant KIX proteins retained their secondary structure, as shown by CD spectra (Fig. S4). Different from the strong impact of Tyr 650 , Ala 654 , and Tyr 658 mutations on c-Myb and pKID binding (21), the Y650A/A654W double mutation lowered the affinity of KIX for BMAL1 by 2-2.5-fold, whereas even less pronounced effects were observed for the Y658A mutant (Fig. 3, C and D). However, comparably moderate effects were reported for mutations in the MLL pocket (I660A/L664A double mutant) or the c-Myb/pKID pocket (Y650A/A654Q/Y658A triple mutant) on the affinity of FOXO3a, which was shown by NMR solution structures to bind to both KIX pockets in two different orientations (29). Based on our interaction studies and reminiscent of FOXO3a, we therefore propose that both KIX-binding pockets play a role in KIX-BMAL1 interactions, and BMAL1 may exploit the KIX pockets in a different manner than MLL1, pKID, and c-Myb.

The acetyl-lysine mimetic BMAL1 K537Q mutation does not affect the BMAL1-KIX binding affinity
BMAL1 acetylation at Lys 537 enhances CRY1 binding to BMAL1 and transcriptional repression of CLOCK:BMAL1 in vivo (13). The acetyl-lysine mimetic K537Q mutation in the purified BMAL1 protein mirrors the in vivo effect of Lys 537 acetylation by increasing the CRY1-binding affinity to the BMAL1-TAD (14,15). As Lys 537 acetylation is daily rhythmic in mouse liver, peaking in the repressive phase (13), we hypothesized that CBP-KIX may preferentially bind to nonacetylated BMAL1 during the transcriptionally active phase. This prompted us to analyze the effect of the acetyl-lysine mimetic BMAL1 K537Q mutation on the binding affinity of KIX to test whether this mutation weakens KIX binding in vitro. However, using NanoSPR and MST, we found that WT BMAL530 and the BMAL530(K537Q) mutant bound to the KIX domain with similar affinities (Fig. 4, A-C). Hence, the K537Q mutation does not affect binding of BMAL1 to the KIX domain, in contrast to the enhanced BMAL1(K537Q)-CRY1 interaction. Likewise, the P624A mutation locking the Trp 623 -Pro 624 imide bond at the BMAL1 C terminus in the trans conformation does not affect BMAL1 affinity for KIX (Fig. 4), as reported previously (16). We propose that the somewhat lower ϳ1 M binding affinity of BMAL530 (WT, P624A, K537Q) to WT KIX determined in this NanoSPR experiment ( Fig. 4) compared with the experiment shown in Fig. 3 is correlated with the fact that now N-terminally His 6 -tagged BMAL1 is used as surface-bound receptor, consistent with the overall lower KIX-binding affinities of N-terminally labeled and His 6 -tagged BMAL1 observed in MST and FP (Fig. 2). Hence, manipulation of the N-terminal end of BMAL530 appears to affect KIX interactions, but this is unrelated to the modification of Lys 537 .

KIX binding does not induce the formation of extended regular secondary structures in BMAL1
The secondary structure of the purified BMAL530 and KIX proteins and their folding behavior upon complex formation were analyzed using CD spectroscopy. The KIX domain was determined to be mostly helical (ϳ58%), consistent with known NMR and crystal structures (20,27,39), whereas BMAL1 was mostly unstructured and intrinsically disordered (ϳ70%) with a local tendency toward ␣-helix ( Fig. 5A and Table 1), which agrees with previous reports (9,14). Interestingly, we did not observe pronounced changes in the secondary structure content of the BMAL530:KIX complex over a course of time rang- were titrated against 50 nM fluorescently labeled His 6 -BMAL530 WT or P624A and K537Q mutants. The normalized fluorescence was plotted against the ligand concentration, and best fits were obtained using a one-site binding model. C, dissociation constants obtained from NanoSPR and MST measurements. The P624A mutation locking the Trp 623 -Pro 624 imide bond in its trans-conformation does not affect the KIX-binding affinity of BMAL1 as described previously (13). The identical affinities of KIX for BMAL530 WT and BMAL1(K537Q) indicated no effect of BMAL1 acetylation on its interaction with the KIX domain.

BMAL1-CBP interactions in the circadian clock
ing from 30 min to overnight (ϳ17 h) after mixing the individually purified KIX and BMAL1 proteins ( Fig. 5B and Table 1). Moreover, the experimental CD spectra of the BMAL530:KIX complex are close to the calculated average of the secondary structure contents of the individual BMAL1 and KIX proteins. Hence, our CD spectra do not suggest the formation of extended regular secondary structures in BMAL1 upon KIX binding, unlike the structuring of pKID and c-Myb induced by KIX binding (20,23). However, based on our CD spectra, we cannot exclude conformational rearrangements associated with minor secondary structure changes (e.g. locally enhanced helix formation). We note that our MST and FP interaction measurements (Figs. 2 and 4B) were performed after a 30-min co-incubation of BMAL1 and KIX at each titrant concentration, and comparable results were obtained after 2-h and overnight co-incubation. Therefore, BMAL1-KIX binding occurs within 30 min, and the lack of pronounced secondary structure changes suggested by our CD spectra is not due to a lack of BMAL1-KIX interactions.

Analyses of the BMAL1:KIX complex architecture by SAXS
To analyze the shapes, dimensions, folding, and conformational flexibility of the individual BMAL1 and KIX proteins and their complexes, we performed in-solution SAXS measurements of our purified BMAL530(P624A) protein, of the KIX domain (KIX(L664C) untethered and 1-10-tethered), and of three BMAL1:KIX complexes including the BMAL1 constructs BMAL1(530 -625)P624A (BMAL530:KIX), BMAL1(517-625) (BMAL517:KIX), or BMAL1(490 -625) (BMAL490:KIX) ( Table  S1). The Guinier analysis of our SAXS data indicated that all of our samples were free of aggregates (Figs. S5B, S6C, and S7B). Furthermore, the estimated molecular weights that were calculated using a Bayesian approach (40) are consistent with a 1:1 stoichiometry of the BMAL1:KIX complexes ( Table 2). The Kratky plot suggests that BMAL530 is mostly unfolded and elongated in shape (Fig.  S6B), whereas the BMAL1:KIX complexes are partially unfolded and elongated in shape (Figs. S6B and S7A). Consistent with our CD spectra ( Fig. 5 and Table 1), we attribute the partial folding of the BMAL1:KIX complexes to contributions of both the helically folded KIX domain and the mostly unstructured BMAL1 fragments. From the distance distribution curves, the maximum particle dimension (D max ) was calculated to be 110 Å for BMAL530 and 64 Å for the tethered and 81 Å for the untethered KIX domain (Fig. 6A, Fig. S5C, and Table 2). These dimensions are consistent with a more elongated shape of BMAL530 compared with the more compact KIX domain and also with the enhanced conformational stability of the 1-10-tethered KIX domain compared with untethered KIX (28). The maximum particle dimensions (D max ) for the BMAL530:KIX, BMAL517:KIX, and BMAL490:KIX complexes were calculated to be 115, 137, and 175 Å, respectively (Figs. 6A and 7A). We attribute the elongated shapes and maximum dimensions (D max ) of the complexes to the BMAL1 proteins, which, as our CD and SAXS data suggest, remain largely unstructured in the complexes and extend beyond the more compact KIX domain.
Based on our SAXS data, we calculated low-resolution models for KIX, BMAL530, and the BMAL530:KIX complex using two programs, DAMMIF/N (Fig. 6, B-D) and GASBOR (not shown), provided within the ATSAS package. The models obtained from both methods reasonably fit the experimental data ( Table 2 and Fig. S6 (D-F)). Next, we overlaid the DAMMIF/N models of BMAL530 and KIX onto the envelope calculated for the BMAL530:KIX complex (Fig. 6E) (correlation values 0.84 and 0.91 for BMAL530 and KIX, respectively). The KIX domain is positioned in a bulkier region in the bottom part of the BMAL530:KIX complex envelope, and the additional, more elongated density emerging on the top is attributed to BMAL1 regions not directly interacting with the KIX domain.
To further investigate the BMAL1 region extending out of the BMAL1:KIX core complex, we also calculated SAXS models for the complexes of the KIX domain with the longer BMAL517 and BMAL490 constructs using DAMMIF/N (Fig.  7B) and GASBOR (not shown), which showed a good fit to the experimental SAXS data (Fig. S7 (C and D) and Table 2). By comparing the SAXS models of the BMAL530:KIX, BMAL517: KIX, and BMAL490:KIX complexes, we found that the longer BMAL1:KIX complexes are indeed more elongated in shape

BMAL1-CBP interactions in the circadian clock
bulkier KIX domain regions and highlights elongated extensions of the complex envelopes on one side of the bulkier region that are significantly longer in BMAL490 than in BMAL530 (Fig. 7C). As the C-terminal BMAL1-TAD residues are known to interact with the KIX domain (9), we assigned the narrowshaped extensions to BMAL1 regions N-terminal to the TAD that do not interact with KIX. Hence, our models identify the location and orientation of the N-terminal ends of the three BMAL1 constructs that emerge from the BMAL1-TAD:CBP-KIX core complex, including the C-terminal BMAL1 residues. To assess on which side of the KIX domain the N-terminal parts of BMAL1 emerge from the KIX:BMAL1-TAD core complex, we positioned high-resolution three-dimensional structures of the KIX domain in the SAXS envelope of 1-10-tethered KIX. The V-shape of the SAXS models as well as the radius of gyration that we obtained for the 1-10-tethered and untethered KIX(L664C) domain from our experimental SAXS data (Fig. S5, Fig. 6C, and Table 2) are in close agreement with the 1-10-KIX crystal structure (PDB entry 4I9O). However, fitting the theoretical scattering data calculated from the 1-10-KIX crystal structure to the experimental SAXS data resulted in overall unsatisfactory fitting values ( 2 ϭ 4.73 (CRYSOL) and 2 ϭ 9.24 (FoXS)) (Fig. S5D). Modeling of N-and C-terminal residues (Gly 586 , Val 587 , Glu 666 -Leu 672 ) that were not included in the 1-10-KIX crystal structure, due to conformational disorder, resulted in a much better fit of the experimental SAXS data ( 2 ϭ 1.51 (CRYSOL) and 2 ϭ 1.49 (FoXS); Fig. S5 (E and F)).
As these residues are included in the NMR structure of the MLL1:KIX complex (PDB entry 2LXS) and, furthermore, the MLL pocket binds the 1-10 compound (27) and is important for BMAL1 interactions (this study), we also used the MLL1: KIX complex structure to position the KIX domain in the 1-10-KIX SAXS envelope (Fig. 7D). The theoretical scattering profile obtained from the NMR structure of KIX (PDB entry 2LXS, eliminating the MLL1 peptide) fits well to the experimental SAXS profile of 1-10-KIX ( 2 ϭ 1.57 after CRYSOL), and the fitting was further improved by model optimization using SREFLEX ( 2 ϭ 1.21 after SREFLEX) (Fig. S5E). The fitting of the KIX NMR structure (model obtained after SREFLEX) and of the 1-10-KIX crystal structure to the SAXS envelope of 1-10tethered KIX and its further superposition onto the SAXS envelope of the BMAL530:KIX complex (Fig. 7D) suggest that the N-terminal BMAL1 region emerges from the KIX domain near the MLL-binding site. By combining our findings with the published literature on BMAL1:KIX interactions (9), we therefore propose that the N-terminal BMAL1 region preceding the

BMAL1-CBP interactions in the circadian clock
BMAL1-TAD, which includes Lys 537 and parts of the BMAL1 G-region, extends out from the BMAL1-TAD:CBP-KIX core complex near the MLL-binding site with the KIX-binding C-terminal BMAL1-TAD region docked in. We note that based on the low-resolution SAXS envelopes, we cannot entirely exclude alternative orientations of the KIX domain within the bulkier regions of the BMAL1:KIX complex envelopes. However, alternative KIX orientations within the bulkier regions fit the complex shapes less well than the orientation presented in Figs. 6 and 7 and Fig. S7. Moreover, they would result in additional extensions of the SAXS envelopes emerging from the KIX:BMAL1-TAD core complex on the opposite side of the N-terminal BMAL1 regions, which is less consistent with the fact that the most C-terminal BMAL1 residues are engaged in KIX binding (9).

Discussion
The C-terminal TAD of BMAL1 and the preceding BMAL1 G-region (Glu 429 -Ile 578 ), including the acetylated Lys 537 , are of critical functional importance for the mammalian circadian clock (8 -13). The BMAL1-TAD facilitates transcriptional repression or activation of CLOCK:BMAL1 by interacting with CRY or the CBP/p300 KIX domain, respectively (9 -12, 14, 15). The CBP-KIX domain, on the other hand, interacts with several transcriptional regulators through its MLL-and cMyb/pKIDbinding pockets (19). To further our mechanistic understanding of the roles of the BMAL1-TAD-CBP-KIX interaction and its interplay with other KIX ligands and with the BMAL1-CRY interaction in circadian regulation, we investigated the contributions of the MLL-and cMyb/pKID-binding pockets of CBP-KIX, of BMAL1 folding, and of BMAL1 Lys 537 acetylation to the BMAL1-KIX interactions. Our quantitative interaction studies (Figs. 2 and 3) revealed that tethering compound 1-10 to KIX(L664C) or KIX(N627C) in the MLL-binding pocket, which significantly weakens the affinity of the MLL1-TAD to KIX (27), only moderately decreases the affinity of BMAL1 to KIX by about 2-fold. The untethered L664C mutation, which leads to a ϳ6-fold reduced MLL1-TAD affinity (27), showed a trend of reducing BMAL1 binding affinity. In contrast, the untethered N627C mutation on the opposite side of the MLL-binding pocket showed a trend of enhancing BMAL1 binding. Altogether, these observations suggest that the MLL-binding pocket is involved in BMAL1 interactions, but with a different binding mode of the BMAL1-TAD compared with the MLL1-TAD, possibly in closer proximity to Asn 627 . Diverse binding modes of different ligands in the MLL pocket have been described structurally and are enabled by the intriguing degree of plasticity of the MLLbinding pocket (29).
Interestingly, a 19-mer MLL1-TAD fragment (MLL19) displaces BMAL1 from the prepurified KIX:BMAL530 complex and is only able to pull down KIX, and not BMAL1, when incubated with the individual BMAL530 and KIX proteins (Fig. 2F). The binding site competition between MLL1 and BMAL1 observed in our pulldown experiment further supports the importance of the MLL-binding pocket for BMAL1-TAD interactions. It remains to be investigated how binding site competition between the BMAL1-TAD and the MLL1-TAD for the CBP-KIX domain would impact the acetylation of MLL1 on Lys 1130 and Lys 1333 by the CBP-HAT domain, which enhances MLL1-dependent H3 Lys 4 trimethylation and subsequent coactivation of CLOCK:BMAL1-dependent circadian genes in vivo (32,33). It is conceivable that recruitment of CBP to the CLOCK:BMAL1 complex requires further stabilization of the BMAL1-CBP interaction (e.g. via additional CBP domains or posttranslational modifications) to enhance specificity of CBP for BMAL1 over alternative KIX-binding transcription factors, such as MLL1. Simultaneous recruitment of CBP and MLL1 to CLOCK:BMAL1 and MLL1 acetylation by CBP would then be possible, as MLL1 interacts with the exon 19 region of CLOCK using a MLL1-binding site near Lys 1130 and Lys 1333 (32,33), whereas CBP is recruited via BMAL1 (9 -12).
To test the role of the second pKID/c-Myb binding pocket in BMAL1 binding, we introduced the mutations Y658A and Y650A/A654W in this pocket, which are known to significantly weaken the interaction of KIX with c-Myb or pKID (21). Interestingly, these mutations only moderately impact BMAL1 binding in our NanoSPR experiments with a ϳ2-2.5-fold weakened affinity for the Y650A/A654W double mutant and an even less pronounced effect of the Y658A mutant (Fig. 3). The moderate impact of alterations in both KIX-binding pockets on BMAL1 affinity is reminiscent of FOXO3a and suggests that the BMAL1-TAD may, like FOXO3a or p53 (29,30), interact with both KIX binding sites, but using binding modes different from those used by MLL1, CREB-pKID, or c-Myb. Solution NMR structures showed that FOXO3a binds to both KIX-binding pockets simultaneously with its conserved regions 2 and 3 (CR2C and CR3) adopting two different conformations and orientations that also deviate from the orientations and interactions of the MLL1-TAD or pKID peptides (29). In the case of the p53-TAD, its two subdomains AD1 and AD2 synergistically bind to both KIX-binding sites (30). Like AD1 and AD2 of the p53-TAD, CR2C and CR3 of FOXO3a, and other transactivation domains of KIX-binding proteins, the BMAL1-TAD contains a ⌽XX⌽⌽ motif (IXXLL 606 ) (where ⌽ represents a hydrophobic residue), which could potentially bind to either of the two KIX-binding sites. Mutational studies showed that the IXXLL 606 motif of BMAL1 is much more important for KIX binding than the seven most C-terminal BMAL1 residues, as the L605A/L606A double mutation completely disrupts the BMAL1-KIX interaction, whereas deletion of the C-terminal BMAL1 residues only leads to a 2-fold reduced affinity (9). As the MLL19-TAD peptide completely displaces BMAL1 (Fig. 2F), we presently favor a scenario in which the more important BMAL1 interaction site is located near the MLL-binding pocket.
Consistent with previous studies (9,14), our CD and SAXS analyses showed that the BMAL1-TAD and the preceding N-terminal regions are intrinsically disordered (Table 1, Fig.  5A, and Fig. S6B), allowing the BMAL1-TAD to dynamically interact with co-regulators in the mammalian circadian clock. The BMAL1-TAD undergoes conformational rearrangements upon binding to CRY1 and KIX, as suggested by manganese-dependent paramagnetic relaxation enhancement studies (9), which prompted us to further biophysically characterize the formation of the BMAL1-TAD:KIX complex by CD spectroscopy. Interestingly, BMAL1-TAD binding to the KIX domain does not lead to pronounced changes in secondary structure

BMAL1-CBP interactions in the circadian clock
content compared with averaged spectra of BMAL530 and KIX proteins (Table 1 and Fig. 5B). Therefore, our CD spectra do not suggest major folding events of extended unstructured BMAL1 regions upon KIX binding, such as the induced folding of the CREB-pKID and cMyb-TAD (20,23). However, the CD data do not exclude conformational rearrangements associated with minor secondary structure changes (e.g. a locally enhanced helical folding of the IXXLL 606 motif region at the KIX interface or orientation changes of the polypeptide chain).
SAXS analyses of the BMAL530 fragment revealed that the BMAL1-TAD and the included parts of the G-region are elongated, lacking a compact domain and mostly unfolded (Fig. 6,  Fig. S6, and Table 2). Furthermore, the extended shape of the BMAL530:KIX complex with a D max similar to unbound BMAL530 and the Kratky plot analyses together suggest that the BMAL1-TAD does not fold upon KIX binding (Fig. 6, Fig.  S6, and Table 2), consistent with our CD spectra ( Fig. 5 and Table 1). The comparison and superposition of the BMAL530: KIX complex with the complexes including longer BMAL1 fragments (BMAL517:KIX and BMAL490:KIX) revealed that parts of the BMAL1 fragments extend out of the bulkier BMAL1-TAD:KIX core complex (Fig. 7, Fig. S7, and Table 2). Based on KIX-BMAL1 interaction studies reported previously (9) and presented herein as well as the fitting of the SAXS envelope and high-resolution three-dimensional structures of the KIX domain into our SAXS models of the three BMAL1:KIX complexes ( Fig. 7D and Fig. S7 (F-H)), we propose that the C-terminal BMAL1-TAD region is bound to the KIX domain utilizing both binding sites, whereas the N-terminal part preceding the BMAL1-TAD and including parts of the G-region emerges from the KIX:BMAL1-TAD core complex near the MLL-binding pocket. This model interpretation is also consistent with our finding that the MLL1-TAD interferes with BMAL1-TAD:KIX complex formation (Fig. 2F). Whereas highresolution structure determination studies are still required to look closely into the structure of the complex, our model provides a plausible starting point to further mechanistically probe BMAL1-CBP interactions.
Using NanoSPR and MST measurements, we found that the K537Q mutation does not affect the affinity of BMAL530 for KIX (Fig. 4). This finding contrasts with the enhancing effect of both the in vitro K537Q mutation (14,15) and the in vivo Lys 537 acetylation (13) on CRY1 binding. Assuming that the K537Q mutation mimics the effect of Lys 537 acetylation on CBP-KIX-BMAL1-TAD interactions as well as that reported for CRY1-BMAL1-TAD interactions (13)(14)(15), our result suggests that Lys 537 acetylation does not affect the BMAL1-TAD-CBP-KIX interaction. However, in the context of full-length CBP and in cells, Lys 537 acetylation could still modulate BMAL1-CBP interactions by alternative mechanisms, as for example in the case of acetylated CREB-pKID (41), and the putative effect of BMAL1 Lys 537 acetylation on CBP-dependent co-activation of CLOCK:BMAL1 or circadian chromatin modification remains to be investigated in vivo.
Unexpectedly, we observed overall lower KIX-binding affinities of BMAL530 in our MST and FP experiments (Fig. 2) compared with our NanoSPR experiments (Fig. 3) and published data (16), which we attribute to the N-terminal fusion of BMAL530 to a fluorescently labeled hexahistidine tag. Therefore, modifications of the BMAL1 region N-terminal to the TAD could have an impact on the BMAL1-TAD-KIX interaction. This finding may be correlated with the documented functional importance of the BMAL1 G-region (9).
Overall, this study has highlighted the significance of the MLL-binding pocket of KIX for BMAL1 interactions and determined the location of the N-terminal regions of our BMAL1 constructs, including parts of the G-region, with respect to the BMAL1-TAD:KIX core complex. Additionally, we observed a moderate effect of mutants in the second KIX-binding pocket, suggesting that it also affects BMAL1 binding. However, our results showed that the second KIX-binding pocket does not play a predominant role for BMAL1 binding, and BMAL1 does not bind to KIX in a CREB-pKID-like manner, in which case a positive cooperative effect of MLL1-TAD binding and 1-10 tethering to Asn 627 might be expected as for pKID (26,28), rather than the negative interference that we observed (Figs. 2 and 3). Based on the presented data, formation of a ternary complex between BMAL1, CBP, and CREB cannot, therefore, be excluded. The interplay between BMAL1 and CREB binding to CBP should be further investigated, as it plays a critical role in defining recruitment of CBP to CLOCK:BMAL1 and/or CREB to regulate circadian transcription and chromatin modification as well as phase resetting and light entrainment of the clock via per1 gene activation (34,35). Thereby, the impact of additional CBP domains also needs to be taken into account as well as posttranslational modifications of pKID and BMAL1, such as CREB-pKID phosphorylation on Ser 133 and Ser 142 , which affects KIX binding and plays a critical role in suprachiasmatic nucleus light entrainment (34). Altogether, our study has significantly advanced our mechanistic understanding of CBP-BMAL1 interactions and lays a stepping stone for further structural and mechanistic analyses of the protein interaction network underlying CBP-dependent gene regulation in the mammalian circadian clock.

BMAL1-CBP interactions in the circadian clock
and resuspended homogenously in lysis buffer containing 50 mM Tris-Cl, pH 7.2, 250 mM NaCl, 5 mM 2-mercaptoethanol, and 10% glycerol. Prior to lysis, 1 mM phenylmethylsulfonyl fluoride and benzonase were added, and the cells were lysed in a fluidizer. The supernatant after the centrifugation was loaded on a Ni 2ϩ affinity column (GE Healthcare), and the protein was eluted with a linear gradient of 0 -500 mM imidazole. The tag was cleaved using His-PreScission protease and dialyzed against a buffer containing no imidazole using a 3-kDa cutoff Roth ZelluTrans dialysis membrane. The cleaved proteins were further cleaned up using a second Ni 2ϩ affinity column for trapping the protease and the tag before loading on a Superdex S75 16/60 column (GE Healthcare). The KIX mutants were purified using the same procedure. The His 6 -tagged BMAL530(K537Q) mutant was expressed at 18°C overnight after induction with 0.5 mM isopropyl ␤-D-1-thiogalactopyranoside and purified as described for WT BMAL1. For the purification of BMAL530(P624A), the protein was further cleaned up using a Q HP anion-exchange column (GE Healthcare) before loading it on a Superdex S75 16/60 column (GE Healthcare). The identity of the proteins was confirmed by MS. The theoretical molecular weights calculated for BMAL1 and KIX proteins are BMAL530 (9.72 kDa), BMAL517 (10.97 kDa), BMAL490 (13.84 kDa), and KIX (10.33 kDa). The BMAL530:KIX, BMAL517:KIX, and BMAL490:KIX complexes were purified by co-lysis of cells overexpressing the individual BMAL530, BMAL517, or BMAL490 proteins for the respective complexes with cells overexpressing the KIX protein. The remaining purification procedure was followed as for the individual BMAL1 and KIX proteins.

Limited proteolysis of the BMAL490:KIX complex
The limited proteolysis experiments were performed by mixing a 40-l protein solution of 1 mg/ml BMAL490:KIX complex with 12 l of 0.1 mg/ml sequencing-grade GluC protease solution. The reactions were incubated for 30 and 60 min. The protease activity was stopped by boiling for 10 min with 5x SDS protein-loading buffer, and the cleavage products were resolved by SDS-PAGE. The N-and C-terminal construct boundaries were determined by performing an in-gel acid hydrolysis using 3 M HCl followed by a peptide cleanup and MS using an LC-MS/MS system (LTQ Orbitrap Velos Pro, Thermo Fisher Scientific).

Confirmation of complete 1-10 compound incorporation by MS
Samples were diluted 10-fold with 30% acetonitrile (LC-MS grade, Roth) and analyzed by direct infusion MS on a Synapt G2-S HDMS mass spectrometer (Waters Corp.). The TOF analyzer of the mass spectrometer was externally calibrated with a sodium formate mixture from m/z 50 to 1990. Spectra were acquired at 25,000 resolution (FWHM) using a 1-s scan time. Mass spectrometric raw data were analyzed manually using MassLynx version 4.1 software (Waters Corp.).

Microscale thermophoresis
The MST measurements were performed in standard treated capillaries on a NanoTemper Monolith NT.115 instrument. The measurements were done following the manual instructions at 90% LED (based on the initial fluorescence intensity) and 60% MST power (IR laser power) with a 30-s laser-on time and 5-s laser-off time. The purified N-terminally His 6 -tagged BMAL530(P624A) protein was labeled with RED-tris-NTA NT-647 dye (Nanotemper Technologies) following the recommended protocol in a buffer containing 25 mM HEPES, pH 7.2, 100 mM NaCl, and 0.05% Tween 20. Briefly, a 200 nM His 6 -BMAL1 protein solution was mixed in a 1:1 ratio (v/v) with a 100 nM RED-tris-NTA dye solution and incubated for 30 min at room temperature. The excess dye was removed by centrifugation at 15,000 ϫ g for 10 min. The unlabeled and untagged KIX domain (WT and mutants) was titrated in the final concentration range of 160 to 0.005 M, mixed in a 1:1 ratio (v/v) with the labeled BMAL1 protein, thereby keeping the final BMAL1 concentration at 50 nM, and filled in the standard capillaries for measurement after incubating for 30 min at room temperature. The normalized fluorescence values were plotted against the untagged protein (titrant) concentration. The data were analyzed using a one-site binding model in GraphPad Prism to obtain the K D values.

Fluorescence polarization
For FP measurements, the purified N-terminally His 6 -tagged BMAL530(P624A) protein was labeled with the RED-tris-NTA NT-647 fluorescent dye (Nanotemper Technologies). The labeling reaction was performed similarly to the MST experiments in a buffer containing 25 mM HEPES, pH 7.2, 100 mM NaCl, and 0.05% Tween 20. Unlabeled proteins (KIX WT/mutants) were titrated in the final concentration range of 0.04 -80 M, thereby keeping the concentration of labeled BMAL1 molecules constant at 50 nM. The reactions were incubated for 30 min at room temperature and transferred to a

BMAL1-CBP interactions in the circadian clock
384-well Corning flat black low-volume plate. The measurements were carried out with a Tecan Spark 20M plate reader in polarization mode at 25°C using excitation and emission wavelengths for the NT-647 dye at 625 and 665 nm, respectively. The dissociation constants were obtained by fitting the concentration-dependent change in polarization using a nonlinear regression function and a one-site specific fit in GraphPad Prism.

Preparation, characterization, and functionalization of gold nanorod particles for NanoSPR
Cetyl trimethylammonium bromide-stabilized gold nanorods were synthesized as described previously (37). Two batches of nanoparticles were prepared that had bulk resonance wavelengths ( res ) centered at 650 nm (batch 1) and 730 nm (batch 2), respectively. Particles from batch 1 had a diameter of 27.2 Ϯ 3.2 nm and a length of 65.4 Ϯ 4.7 nm, whereas particles from batch 2 had a diameter of 33.2 Ϯ 3.7 nm and a length of 91.4 Ϯ 7.0 nm, as determined by transmission EM. Both particle batches led to similar signal/noise ratios, and their application in NanoSPR experiments resulted in similar K D values. To enable binding to the His-tagged protein receptors, the cetyl trimethylammonium bromide layer around the particles was replaced by NTA, and the His tags of the receptor proteins were complexed with nickel, adapting an established protocol (38).

Single-nanoparticle plasmon spectroscopy and NanoSPR
We flushed and sequentially immobilized the different batches of functionalized nanoparticle receptor conjugates in the microfluidic flow cell and recorded the position of every single nanosensor as described (37) to create a position-encoded sensor substrate. Before and after each particle deposition step, we washed the flow cell generously with buffer (25 mM HEPES, pH 7.2, 100 mM NaCl, 5% glycerol) to remove unbound particles and proteins.
The optical dark-field spectroscopy setup used to record single-particle spectra was described in detail recently (36,44). Each nanoparticle receptor sensor has a characteristic plasmon resonance wavelength ( res ), which will have a spectral shift (⌬ res ) upon binding of biomolecules near the nanoparticle surface (37). To obtain satisfactory statistics of the responses over many sensors, we used a spectral imaging detection method that allowed for the investigation of all of the particles within the field of view simultaneously as described (36,44).
To carry out a titration experiment, 100 l of a ligand solution (untagged KIX or BMAL530(P624A) proteins) with a given concentration was flushed into the flow cell at a flow rate of 50 l/min, and the spectral shift (⌬ res ) of every particle (80 -200 particles for each protein receptor) was recorded with a time resolution of about 1 min at this ligand concentration until an equilibrium value (⌬ max ) was reached. This process was repeated for 8 -10 different ligand concentrations (0.045-25 M). To determine the binding parameters, the following equation was used, (Eq. 1) (with the coverage ⍜ ϭ ⌬/⌬ max ) as fitting function, and the dissociation constants were obtained using a one-site binding model in GraphPad Prism by plotting the normalized wavelength shifts as a function of ligand concentration.

GST pulldown
The His 6 GST-tagged human MLL1-TAD peptide (denoted as MLL19) comprising residues 2840 DCGNILPSDIMDFV-LKNTP 2858 (equivalent to Asp 2836 -Pro 2854 with identical sequence in mouse MLL1) was subcloned from full-length human MLL1 (in pTARGET TM ) into pCoofy3, expressed in E. coli, and purified using Ni 2ϩ -affinity and size-exclusion chromatography in a buffer containing 25 mM HEPES, pH 7.2, 150 mM NaCl, 5% glycerol, and 2 mM DTT. A concentration of 4 M GST-tagged MLL19 (bait), 20 M BMAL530:KIX complex, or individual BMAL530 and KIX proteins (prey) and ϳ20 l of GSH agarose (Sigma) was used for pulldown in a reaction volume of 200 l. The reactions were washed three times with the above buffer following the overnight rotation at 4°C. The protein was eluted with SDS protein-loading buffer and boiled for ϳ5 min followed by SDS-PAGE.

CD spectroscopy
CD spectra were recorded with a Jasco-815 spectropolarimeter using a quartz cuvette of 0.1-cm pathlength. The measurements were performed in continuous scan mode with accumulation of six scans for each measurement at 4°C for BMAL530 and KIX (WT and mutants) and at 20°C for the complexes. The data were collected at every nm from 185 to 250 nm at a scan speed of 50 nm/min. The purified proteins were diluted to a final concentration of 10 -20 M in 20 mM sodium phosphate buffer, pH 7.2. The CD spectra from individual BMAL530 and KIX proteins were averaged for the calculated spectra of the complex. The machine units (in millidegrees, ) were converted to mean residue ellipticity (MRE, [] R ) from mean residue weight (MRW), path length (l, in cm), and protein concentration (c, in mg/ml) using the following equation. The data were deconvoluted, and the protein secondary structure was calculated using the CONTIN algorithm in the DichroWeb server (45) with the reference data set 6 (46).

SAXS
The data were collected on the SAXS beamline P12 (47) at PETRA III, DESY, Hamburg with an X-ray wavelength of 1.24

BMAL1-CBP interactions in the circadian clock
Å. The measurements of protein solutions were carried out in the concentration range of 0.1-11 mg/ml to check for the interference between protein molecules at different concentrations and to obtain high-quality data. The PRIMUS Qt program in the ATSAS 2.8.4 software package (48) was used for processing the data, and from small q-regions of the scattering data, the radius of gyration (R g ) was determined using the Guinier approximation. The molecular weights were determined using a Bayesian inference approach (40) implemented in the ATSAS 2.8.4. software. The pairwise-distribution function (P(r)) was computed using GNOM (49), which gives the maximum dimension (D max ) of the macromolecule. The GNOM output file was used by the ab initio shape-determination programs GASBOR (50) and DAMMIF (51), which generated 10 independent low-resolution envelopes that were averaged together using DAMAVER (52) and refined using DAMMIN (53).
The programs FoXS (54) and CRYSOL (55) were used to calculate the theoretical scattering profiles from published KIX structures (PDB entries 4I9O and 2LXS) and for fitting them to the experimental SAXS data. The SREFLEX program (56) was used to further improve the agreement of the NMR structure of the KIX domain (from PDB entry 2LXS, eliminating MLL1 peptide) to the experimental SAXS data calculated for 1-10-KIX by estimating flexibility of the high-resolution structure. To improve the fit of the 1-10-KIX crystal structure to the SAXS envelope, residues Gly 586 , Val 587 , and Glu 666 -Leu 672 , which were missing in the crystal structure of 1-10 KIX(L664C) (PDB entry 4I9O), were modeled using MODELLER version 9.20 (57). The program Chimera (58) was used for the conversion of SAXS envelopes into maps (command: molmap) and to superpose the maps of the KIX:BMAL530, KIX:BMAL517, and KIX: BMAL490 complexes onto each other (tool: "Fit in Map"). The Fit in Map option of Chimera was also used to fit the SAXS envelopes obtained for KIX and BMAL530 into the SAXS envelope of the BMAL530:KIX complex and to fit the KIX structures into the SAXS map of 1-10-KIX.

Data accession
The SAXS scattering data collected at the P12 beamline has been deposited in the Small Angle Scattering Biological Data Bank (SASBDB) (59)