A slow conformational switch in the BMAL1 transactivation domain modulates circadian rhythms

SUMMARY

The C-terminal transactivation domain (TAD) of BMAL1 (Brain and muscle ARNT-like 1) is a regulatory hub for transcriptional coactivators and repressors that compete for binding and consequently contributes to period determination of the mammalian circadian clock. Here, we report the discovery of two distinct conformational states that slowly exchange within the dynamic TAD to control timing. This binary switch results from cis/trans isomerization about a highly conserved Trp-Pro imide bond in a region of the TAD that is required for normal circadian timekeeping. Both cis and trans isomers interact with transcriptional regulators, suggesting that isomerization could serve a role in assembling regulatory complexes in vivo. Towards this end, we show that locking the switch into the trans isomer leads to shortened circadian periods. Furthermore, isomerization is regulated by the cyclophilin family of peptidyl-prolyl isomerases, highlighting the potential for regulation of BMAL1 protein dynamics in period determination.

TOC Image

Mammalian circadian clocks are intrinsic molecular timekeeping systems that coordinate physiological processes with external environmental cues in order to appropriately time daily activities. This coordination is achieved by two interlocked transcription feedback loops that control the temporal basis of expression for over 40% of the mammalian genome (Zhang et al., 2014). The heterodimeric transcription factor CLOCK:BMAL1 sits at the core of the primary feedback loop that directs the chronometric transcription of clock-controlled genes. Circadian timekeeping is established through the coordinate regulation of CLOCK:BMAL1 by transcriptional coactivators, such as CBP/p300, and the dedicated circadian repressors PER and CRY (Partch et al., 2014). Many processes contribute to the timing of activation and repression, such as the localization, phosphorylation, and degradation of repressors in the nucleus (reviewed in (Gallego and Virshup, 2007)). However, the mechanisms that regulate the changing architecture of CLOCK:BMAL1 transcriptional regulatory complexes throughout the day are still poorly understood (Koike et al., 2012). The highly dynamic BMAL1 transactivation domain (TAD) is one hub for these interactions, as it interacts with both coactivators and repressors and is necessary for circadian oscillations (Kiyohara et al., 2006; Park et al., 2015; Xu et al., 2015). Modulating affinity of the BMAL1 TAD for its regulators elicits large changes in period, demonstrating an important role in timekeeping of the transcription-based clock (Xu et al., 2015). Identification of processes that fine-tune interactions between the BMAL1 TAD and transcriptional regulators will shed light on the mechanism by which animals measure time and use it to coordinate behavior and physiology with the environment.

By inducing changes in conformation, protein-protein interactions, or subcellular localization, post-translational modifications play critical roles in controlling signaling pathways and information processing. Additionally, proteins exhibit a range of dynamic behaviors on different timescales that can also contribute to functional output (Henzler-Wildman and Kern, 2007). From the fast, stochastic motions of intrinsically disordered regions to the longer timescales of coordinated domain motions and protein folding, the dynamic behavior of proteins ultimately governs how they regulate biological processes. One such behavior is conformational cis/trans isomerization about a proline-containing imide peptide bond (Xaa-Pro); popularly dubbed a molecular timer (Lu et al., 2007), isomerization is an intrinsically slow process (milliseconds to minutes) that can be enzymatically accelerated by peptidyl prolyl isomerases (PPIases) by up to ~4–5 orders of magnitude (Schmid, 1993). Proline isomerization is relatively rare, as only ~6% of imide bonds are estimated to undergo isomerization (Stewart et al., 1990), but it has a profound impact on diverse cellular processes, such as transcription (Bataille et al., 2012; Nelson et al., 2006), protein folding (Wedemeyer et al., 2002), ion channel gating (Lummis et al., 2005), and protein degradation (Liou et al., 2011). Likewise, proline isomerases are pivotal components of many intracellular signaling pathways through their acceleration of this otherwise slow process (Brazin et al., 2002; Lang et al., 1987; Saleh et al., 2016). Dysfunctional regulation of proline isomerization and/or PPIase activity has been implicated in cancer (Zhou and Lu, 2016), Alzheimer’s disease (Nakamura et al., 2012) and regulation of circadian rhythms in Drosophila (Kang et al., 2015).

Here, we report the discovery of a slow conformational switch in the BMAL1 TAD that modulates mammalian circadian rhythms. Using NMR spectroscopy, we show that cis/trans isomerization about a conserved Trp-Pro imide bond generates this conformational exchange, which we have dubbed the ‘TAD switch’. To study the roles of individual isomers in interactions with circadian transcriptional regulators, we developed locked cis and trans isomers using site-directed mutagenesis or peptide synthesis with unnatural amino acids. Locked isomers bind CRY1 and CBP KIX with similar affinities, yet locking the TAD into its trans isomer shortens the circadian period in cell-based assays, demonstrating that exchange between these two conformations contributes to circadian timekeeping. Using NMR, we determined that the timescale of isomerization is intrinsically slow, taking minutes to complete a cycle of exchange. We then identified a group of PPIases within the cyclophilin family that significantly enhance rates of isomerization. Inhibition of cyclophilins lengthens circadian period in a switch-dependent manner, suggesting that enzymatic modulation of intrinsically slow dynamics at the BMAL1 TAD could play a role in tuning the circadian period in vivo.

RESULTS

PROLINE ISOMERIZATION ACTS AS A MOLECULAR SWITCH IN THE C-TERMINUS OF BMAL1

The BMAL1 TAD acts as a regulatory hub that interacts with positive or negative transcriptional regulators as a function of circadian time (CT) to control the activation state of CLOCK:BMAL1 (Koike et al., 2012). We previously used NMR spectroscopy to identify overlapping binding sites of the transcriptional coactivator CBP/p300 and the repressor CRY that map to two distinct sites in the BMAL1 TAD: the predicted alpha helical region in the center of the TAD and the extreme C-terminus (Figure 1A) (Xu et al., 2015). ¹⁵N-¹H heteronuclear single quantum coherence (HSQC) NMR spectra display a peak for each of the constituent N-H bonds in the disordered TAD. The chemical shift, or location of the peaks, represents a population-weighted average of conformations that interconvert in fast exchange. To our surprise, the ¹⁵N-¹H HSQC spectrum of the BMAL1 TAD revealed two distinct resonances for each of the 8 C-terminal residues (Figures 1B and S1A), indicating slow exchange between two conformations localized to this distal site. We performed liquid chromatography/mass spectrometry analysis on the ¹⁵N-labeled NMR sample, which demonstrated the presence of a single, highly pure peptide of the expected molecular weight (Figures S1B–C). Moreover, truncation of the C-terminal 7 residues resulted in a ¹⁵N-¹H HSQC spectrum devoid of peak doubling, confirming that chemical exchange requires the extreme C-terminus of the TAD (Figure S1A).

Figure 1. Isomerization about a conserved Trp-Pro imide bond in the BMAL1 C-terminal TAD.

(A) Domain schematic of mouse BMAL1 showing the chemical shifts (Δδ) in the TAD after binding CRY1 CC helix (purple) or CBP KIX domain (green). *, proline residues that lack crosspeaks in ¹⁵N-¹H HSQC NMR spectra. (B) Selected regions of ¹⁵N-¹H HSQC spectra of ¹⁵N BMAL1 TAD displaying backbone amide peaks for two isomers at W624 and L626. (C) Cross peaks for P623 and P625 C_β and C_γ atoms are shown in strips from the ¹⁵N-edited (H)C(CO)NH-TOCSY of ¹³C/¹⁵N BMAL1 TAD at ¹⁵N planes for W624 and L626 amides. Average ¹³C shifts for trans (gray) and cis (blue) isomers from (Shen and Bax, 2010). (D) The W624-P625 imide bond in cis and trans conformations. (E) Sequence alignment BMAL1, BMAL2 and CYCLE from insects with a vertebrate-like clock (iBMAL1). (F) Regions of ¹⁵N-¹H HSQC spectra of 8-mer switch peptides from mouse (FSDLPWPL, black) and dwarf honey bee (FSGLPWPLP, peach) showing cis and trans peaks for W624 indole. See also Figure S1.

To identify the structural basis for this conformational heterogeneity, we turned to the C(CO)NH TOCSY NMR experiment, which correlates sidechain carbon chemical shifts with the following amide peptide bond to provide sequence-specific information about the local environment. Looking back from the doubled amide peaks for residue L626, the chemical shifts of P625 ¹³C_β and ¹³C_γ atoms unambiguously identified that the W624–P625 imide bond is found in two distinct conformations, a cis and a trans form based on a comparison to NMR chemical shift databases (Figures 1C, D) (Shen and Bax, 2010). No cis isomer was detected for P623 or any of the other three imide bonds in the BMAL1 TAD construct (Figures 1C and S1D–F), demonstrating that peak doubling in the C-terminus is due solely to slow isomerization of the Trp-Pro bond. Using the relative abundance of several representative peaks of the two isomers, we determined that the population of the TAD switch under these conditions is approximately 65% trans and 35% cis isomers (Figure S1G).

CONSERVATION OF THE TAD SWITCH FROM INSECTS TO VERTEBRATES

To begin exploring the functional significance of the TAD switch, we first examined its conservation across BMAL orthologs (Figure 1E). We noted that the proline of the switch is not conserved in vertebrate BMAL2, a homolog of BMAL1 that has an active TAD but cannot sustain circadian cycling outside of the suprachiasmatic nucleus (Shi et al., 2010; Xu et al., 2015). Phylogenetic analysis of CYCLE, the insect ortholog of BMAL1, demonstrates its divergence into two distinct gene families: a Drosophila-like CYCLE (dCYC) that possesses only the N-terminal bHLH, PAS-A, and PAS-B domains, and a vertebrate BMAL1-like CYCLE that also contains the C-terminal TAD (Chang et al., 2003). The BMAL1-like CYCLE genes found in insects also possess a vertebrate-like cryptochrome with transcriptional repressor activity, suggesting that the network architecture of these molecular clocks is similar to vertebrates (Rubin et al., 2006; Zhu et al., 2008). Because these insect CYC genes exhibit higher functional and structural homology to vertebrate BMAL1 than Drosophila CYC, we hereafter refer to these genes as insect-BMAL1 (iBMAL1). We found that the TAD switch is strictly conserved in all iBMAL1 genes, indicating that the presence of the switch in BMAL1 predates the divergence of insects and vertebrates about 600 million years ago (Peterson et al., 2004).

Although the eight residues of the BMAL1 TAD switch are highly conserved throughout metazoans, we noted several substitutions upstream of the conserved Trp-Pro switch and the inclusion of an additional C-terminal proline in some invertebrates (Figure 1E). To determine if these sequence variations affect the switch, we synthesized 8-mer peptides using the vertebrate BMAL1 and Apis florea iBMAL1 sequences and analyzed peak intensities from ¹⁵N-¹H HSQC spectra collected on natural abundance samples. The vertebrate switch peptide (FSDLPWPL) displayed an equilibrium population of isomers similar to the intact mouse BMAL1 TAD (Figures 1F and S1G–H), as did the iBMAL1 switch peptide (FSGLPWPLP) (Figure 1F), suggesting that insects with a vertebrate-like clock likely share switch functionality with mouse BMAL1.

TRP AND PRO ARE THE KEY RESIDUES COMPRISING THE TAD SWITCH

To probe the importance of the two switch isomers on circadian rhythms, we first set out to identify local sequence requirements that contribute to the TAD switch. By carefully defining these local factors in vitro, we aimed to validate a set of molecular tools that would allow us to explore switch function in the cellular environment. With rare exception, only proline allows the cis isomer to arise in a peptide bond (Pal and Chakrabarti, 1999). As expected, mutation of P625 to Ala eliminated detection of the cis isomer by NMR, providing us with a trans-locked BMAL1 (Figures 2A, B). Conversely, integration of the bulky analog 5,5-dimethyl Proline (dmP) (Lummis et al., 2005; Nakamura et al., 2012) in place of P625 produced the opposite effect, a TAD switch exclusively populating the cis isomer (Figures 2A, B). The backbone geometry of these mutants was verified using ¹³C-¹H HSQC and ¹H-¹H TOCSY spectra collected on natural abundance peptides (Figures S2A, B). The P625A trans-locked switch mutant was also incorporated into the intact BMAL1 TAD (residues 579–626) to establish that cis/trans isomer ratios were similar by NMR spectroscopy (Figure S2C).

Figure 2. Locked mutants of the TAD switch shorten the circadian period.

(A) Representation of cis content of 8-mer TAD switch peptides for P625 and W624 mutants compared to the intact ¹⁵N BMAL1 TAD. Cis content was calculated from peak volumes of residues 624 and 626 in ¹⁵N-¹H HSQC and ¹H-¹H TOCSY NMR spectra. (B) ¹H NMR spectra from FSDLPWPL (black), FSDLPWAL (red) and FSDLPWdmPL (blue) 8-mer TAD switch peptides highlighting the W624 indole region. (C) Synchronized circadian bioluminescence records from Bmal1^−/−; Per2^Luc mouse fibroblasts complemented with WT (gray), W624A (pink), P625A (red), or Δswitch (619X, green) Bmal1. Black line, mean luminescence ± s.d. from n = 6–8 replicates from 2 independent clonal lines in indicated colors. D) Circadian period of complemented fibroblast lines from panel (C). Individual period measurements with mean ± s.d. ***P<0.01 and ****P<0.0001 compared to WT Bmal1 by two-tailed t test. See also Figure S2.

Long range interactions affecting the equilibrium population of cis/trans isomers have been reported in some highly structured systems (Wedemeyer et al., 2002), but for intrinsically disordered regions such as the BMAL1 TAD, long range constraints are unlikely to affect isomerization (Theillet et al., 2014). By contrast, the identity of the (i-1) residue that precedes the proline has a profound impact on the equilibrium population of cis isomers. Aromatic amino acids in the (i-1) position increase the propensity of an imide bond to sample the cis conformation through stabilizing interactions of the π aromatic face with the proline C–H_α bond (Zondlo, 2013), while small electron-poor amino acids typically decrease stability of the cis isomer (Reimer et al., 1998; Shen and Bax, 2010). As predicted, the replacement of W624 with Ala resulted in no observable cis isomer population in both the 8-mer peptide and the intact TAD (Figures 2A, B and S2A–C). Likewise, decreasing the aromaticity of the (i-1) residue by substitution of W624 to Tyr resulted in a decrease in the cis population (Figures 2A and S2A, B). Altogether, these data demonstrate that both Trp and Pro residues are important for the observed conformational switch in the BMAL1 TAD.

THE TRANS-LOCKED TAD SWITCH DRIVES SHORTENED CIRCADIAN RHYTHMS

Mutations in the BMAL1 TAD elicit control over the period of the transcription-based clock by differentially regulating affinity for transcriptional coactivators and repressors. For example, substitution of two key residues in the central alpha helical region of the TAD (L606A/L607A) disrupted circadian rhythms altogether by eliminating interactions with regulators, while other mutations led to circadian periods from ~19 to 26 hours; notably, deletion of the last 7 residues of the TAD (619X, referred to as Δswitch here) shortened the intrinsic period by approximately 3 hours (Xu et al., 2015). To examine the functional consequences of disrupting the TAD switch on circadian rhythms, we stably incorporated different trans-locked Bmal1 mutants into Bmal1^−/−;Per2^Luc cells. In our hands, complementation with wild type Bmal1 resulted in a period of approximately 23 hours (Figures 2C, D and S2D–E), while the P625A mutant decreased this period by over 1 hour. Complementation with W624A Bmal1 led to a ~3 hour decrease in period, on par with the shortened period observed upon deletion of the entire switch region (Figures 2C, D and S2D–E) (Xu et al., 2015). Moreover, the W624A/P625A double mutant exhibited a short period similar to the W624A mutant (Figure S2D–E), demonstrating that while locking the switch into the trans isomer has a significant effect on period, this effect is further enhanced by deletion of the bulky aromatic sidechain of W624.

BOTH ISOMERS OF THE TAD SWITCH INTERACT WITH TRANSCRIPTIONAL REGULTORS

To begin to probe the molecular mechanism for the short period phenotype observed upon complementation with trans-locked mutants of Bmal1, we quantitatively analyzed the interaction of cis and trans isomers of the TAD with known transcriptional regulators using NMR and fluorescence anisotropy. NMR studies performed on the wild-type ¹⁵N BMAL1 TAD demonstrated that the CC helix and KIX domains of regulators CRY1 and CBP, respectively, each interact with both isomers present in the native TAD (Figure 3A, left panels). The same ¹⁵N-¹H HSQC titration experiment was also performed on the trans-locked ¹⁵N BMAL1 TAD P625A mutant, giving rise to chemical shift changes that were highly similar to those of the native trans isomer (Figures 3A, right panels, and Figure S3A–C), suggesting that the P625A mutant is a reasonable proxy for the native trans isomer of the TAD.

Figure 3. Both isomers of the TAD switch interact with transcriptional activators and repressors.

(A) Regions of the ¹⁵N-¹H HSQC spectra showing 100 μM WT (left panels) or P625A (right panels) ¹⁵N BMAL1 TAD upon titration of CRY1 CC (purple) or CBP KIX (green), with increasing concentrations from 25–200 μM indicated by darker colors. (B) Fluorescence anisotropy data for CRY1 PHR (left) and CBP KIX (right) with wild type (black), P625A (red), P625dmP (blue) and Δswitch (green) 5,6-TAMRA-labeled short BMAL1 TAD (residues 594–626). Mean polarization data from one representative experiment (n = 4 replicates) of 3 independent assays. See also Figure S3.

We then determined affinities of locked variants of the TAD switch for CRY1 and the CBP KIX domain using fluorescence anisotropy with short labeled BMAL1 TAD peptides encompassing the highly conserved region from residues 594–626 (Xu et al., 2015). To promote robust changes in fluorescence polarization from the short TADs, we elected to use the ~50kDa CRY1 photolyase homology region (PHR) that contains the CC helix instead of the isolated CC peptide because they have similar affinity for the TAD (Czarna et al., 2013; Xu et al., 2015). In this assay, both WT and Δswitch TADs bound the CRY1 PHR and CBP KIX domain with affinities similar to those previously determined by isothermal titration calorimetry (Figure 3B and Table 1) (Czarna et al., 2013; Xu et al., 2015). Moreover, analysis of the CRY1 binding curve suggested positive cooperativity with a Hill coefficient of 1.6, possibly arising from enforced proximity effects (Ferrell and Cimprich, 2003; Pullen and Bolon, 2011) of the two CRY1 binding motifs at the central alpha helix and switch region of the TAD. In support of this, truncation of the switch region eliminated the apparent cooperativity (Table 1), further demonstrating its importance for interactions with regulators. However, binding assays with cis and trans-locked short TADs demonstrated that affinities for the CRY1 PHR and CBP KIX domain were similar to wild-type (Figure 3B, Table 1 and Figure S3D–E), indicating that effects of the TAD switch on circadian period appear not to arise from differential affinity for either of the known binding partners under these conditions.

Table 1.

Affinity of BMAL1 TAD mutants and conformationally-locked isomers for transcriptional regulators

	CRY1 PHR		CBP KIX
BMAL1 TAD	KD (μM)	Hill	KD (μM)	Hill
Wild Type	0.90 ± 0.27	1.6 ± 0.1	1.34 ± 0.40	0.99 ± 0.13
P625A	0.99 ± 0.27	1.8 ± 0.1	1.59 ± 0.66	0.98 ± 0.28
P625dmP	0.86 ± 0.17	1.9 ± 0.1	1.16 ± 0.44	1.02 ± 0.48
Δswitch	4.28 ± 0.33	0.9 ± 0.04	4.63 ± 1.05	0.91 ± 0.19

Data were acquired by fluorescence polarization assay and fit to a one-site binding model with Prism 6.0. Values shown are averages of three or four independent experiments ± s.d. with n = 2 or 3 replicates each.

ISOMERIZATION OF THE TAD SWITCH OCCURS ON THE TIMESCALE OF MINUTES

The ability to switch between distinct protein conformations can regulate signaling by modulating affinity or occluding binding interfaces (Brazin et al., 2002; Sarkar et al., 2007), or by exerting kinetic control over isoenergetic conformational states to influence the selection of partners (Phillips et al., 2013). Isomerization about the imide bond is an inherently slow process, leading to experimentally determined rates of proline isomerization from milliseconds to nearly an hour in vitro (Eckert et al., 2005; Grathwohl and Wuthrich, 1981), considerably longer than the nanosecond backbone dynamics typically encountered in disordered proteins (Henzler-Wildman and Kern, 2007). To determine the timescale of exchange between cis and trans isomers in the BMAL1 TAD, we used ¹⁵N-¹H ZZ exchange NMR experiments that incorporate a short delay after encoding the ¹⁵N frequency to capture interconversion via formation of cross peaks that align with the ¹H frequency of the new state. When performed at 25°C, the absence of characteristic cross peaks in the ¹⁵N-¹H ZZ exchange assay (Figure 4A) demonstrated that isomerization was too slow to be detected at room temperature. Because intrinsically disordered proteins like the BMAL1 TAD already lack structural elements that are typically disrupted by high temperature, we increased the temperature to enhance isomerization rates. First, we collected a series of ¹⁵N-¹H HSQC spectra at increasing temperatures up to 70°C (Figures 4B and S4A), observing that integrity of the BMAL1 TAD was retained after incubation at temperatures up to 70°C, with an essentially identical ¹⁵N-¹H HSQC spectrum upon return to 25°C (Figure S4B), indicating no lasting effect of high temperature on the protein.

Figure 4. Slow isomerization of the TAD switch occurs on the timescale of minutes.

(A) Highlighted region of the 2D spectrum from a ¹⁵N-¹H ZZ-exchange assay performed at 25°C of ¹⁵N BMAL1 TAD displaying cis (dark green), trans (dark blue), cis to trans (light green) and trans to cis (light blue) cross peaks for L626 at a delay = 1 s. Dashed circles, location of exchange cross peaks. (B) Overlay of ¹⁵N-¹H HSQC spectra showing the cis and trans peaks of L626 at increasing temperatures. (C) Snapshot of ZZ-exchange assays at delay = 1 s at indicated temperatures. (D) Eyring analysis of exchange rates vs. temperature from ZZ-exchange assays with errors displayed in s.d. (E) Free energy plot showing the calculated activation energy of isomerization for the BMAL1 TAD. See also Figure S4.

To explore the effect of temperature on switch kinetics, we collected ¹⁵N-¹H ZZ exchange datasets at increasing temperatures until we observed cross peaks indicative of exchange. These cross peaks first appeared at 55°C and grew in intensity with increasing temperature (Figure 4C). Intensities for peaks representing cis and trans isomers of the W624 side chain indole and L626 backbone amide, as well as exchange cross peaks, were extracted and plotted as a function of delay time (Figure S4C). Exchange rates were calculated independently for temperatures from 55–65°C using the Bloch-McConnell equations (Farrow et al., 1994). We then plotted temperature-dependent exchange rates using the Eyring equation to extrapolate the kinetics of exchange to temperatures lower than 55°C (Figure 4C and Supplemental Table 1). At 25°C, the time for cis to trans isomerization is calculated to take 3.64 minutes, while trans to cis isomerization takes 6.30 minutes, for an overall exchange lifetime of approximately 10 minutes. Based upon these data, we calculated a barrier for cis/trans isomerization of ~20.4 kcal/mol, with a difference in stability between the two isomers of ~0.5 kcal/mol (Figure 4D), both on par with values for analogous systems (Reimer et al., 1998). Compared to the fast motions of the intrinsically disordered TAD backbone suggested by predominantly negative heteronuclear ¹⁵N-¹H NOE values (Figure S4D), slow isomerization of the Trp-Pro imide bond suggests that functionally relevant conformational dynamics may exist over at least twelve orders of magnitude in timescale in the BMAL1 TAD (Figure S4E).

CYCLOPHILINS CAN ACCELERATE INTERCONVERSION OF THE TAD SWITCH

Intrinsically slow rates of cis/trans isomerization are often enhanced catalytically by peptidyl-prolyl isomerases to regulate cellular signaling events. Notably, peptidyl-prolyl isomerases regulate circadian rhythms in Drosophila; however, activity of the Pin1-like Dodo isomerase is phosphoserine-specific and appears to target PER proteins (Kang et al., 2015). Given that the BMAL1 TAD doesn’t possess this phosphospecific consensus motif, we chose to focus our initial attention on cyclophilins for two reasons. First, the broadly expressed peptidyl prolyl isomerase A (PPIA; also known as Cyclophilin A, CypA) co-immunoprecipitated with BMAL1 in a screen for interacting partners (Lipton et al., 2015), and second, the cyclophilin family has eight isoforms with nuclear localization (Adams et al., 2015) where we believe regulation of the TAD is likely to occur. To determine if PPIA could regulate the TAD switch in vitro, we performed a ¹⁵N-¹H ZZ exchange NMR experiment on the BMAL1 TAD at room temperature in the presence of sub-stoichiometric concentrations of PPIA. As cyclophilins have notoriously low K_ms (0.3–1 mM) for their substrates (Coelmont et al., 2010; Schmid, 1993) this precluded a traditional analysis of the catalytic efficiency of PPIA for the TAD. However, we found that PPIA increased the relative rate of isomerization at room temperature by ~300-fold (Figure 5A).

Figure 5. Cyclophilins accelerate isomerization of the TAD switch to regulate circadian period.

(A) Domain schematics for cyclophilins tested against the BMAL1 TAD and relative rate enhancement compared to uncatalyzed isomerization at room temperature. (B) Highlighted region of ¹⁵N-¹H ZZ-exchange spectra displaying the W624 indole of the ¹⁵N TAD with indicated cyclophilins. Spectra from a ZZ-exchange time delay series (delay = 0–1 s) are overlaid in sequentially darker colors. (C) Synchronized circadian bioluminescence records from U2OS Bmal1-dLuc fibroblasts in the presence of DMSO, Cyclosporin A (CsA), or Deltamethrin. One individual trace shown for clarity from 2 independent assays with n = 3 replicates. (D) Mean circadian period of cultures upon treatment with DMSO (black), CsA (green) or Deltamethrin (purple) ± s.d. from n = 3 replicates with 2 independent assays. (E) Comparison of mean period changes ± s.d. as a function of CsA concentration in Bmal1^−/−; Per2^Luc fibroblasts complemented with WT (black), P625A (red), or W624A/P625A (gray) Bmal1. ***P<0.01 compared to WT Bmal1 by two-tailed t test. See also Figure S5.

We subsequently purified and assayed each of the predominantly nuclear cyclophilins PPIE, PPIG, PPIH, PPIL1, PPIL2, PPIL3, PPWD and CWC27 for activity against the TAD switch by acquiring ¹⁵N-¹H ZZ exchange data under similar conditions at room temperature. We found that the cyclophilins PPIE, PPIG, PPIH, PPIL1, PPIL3 all significantly increased the rate of isomerization in BMAL1, although their acceleration of exchange rates in the TAD varied by over 10-fold (Figures 5A, B). By contrast, the isomerases PPIL2, PPWD, and CWC27 were either not active on the TAD or had activity that was below the detection limit of our ¹⁵N-¹H ZZ exchange assay. Both PPIL2 and CWC27 were previously shown to lack activity on generic peptide substrates used to assay cyclophilin activity, suggesting they are catalytically dead (Davis et al., 2010). We then explored if cyclophilin activity on the TAD was restricted to isoforms with nuclear localization by testing the activity of PPIF (also known as CypD), which is localized exclusively in mitochondria (Lin and Lechleiter, 2002). We found that PPIF was at least as active on the TAD as the top-ranked nuclear cyclophilins in our assay (Figures 5A, B), indicating that cyclophilins can exhibit robust activity on the BMAL1 TAD when removed from a cellular context that might otherwise impart substrate selectivity. To date, very little is known about factors that dictate the substrate selectivity or activity of cyclophilins, although they share relatively similar active sites (Davis et al., 2010). Notably, some of the cyclophilins that are active on the BMAL1 TAD in vitro are also expressed in a circadian manner in vivo (Figure S5, (Pizarro et al., 2013)). Therefore, the selectivity and/or activity of cyclophilins on the TAD may be conferred in vivo by regulated changes in abundance, subcellular localization or through formation of different regulatory complexes with CLOCK:BMAL1 throughout the day (Koike et al., 2012; Lee et al., 2001).

INHIBITION OF CYCLOPHILINS INCREASES CIRCADIAN PERIOD LENGTH

To explore if cyclophilins influence timekeeping by the mammalian circadian clock, we treated human U2OS osteosarcoma cells stably transfected with a Bmal1-dLuc bioluminescent circadian reporter (Vollmers et al., 2008) with the broad specificity cyclophilin inhibitor, Cyclosporin A (CsA). This cell permeable cyclic peptide binds to the active site of the cyclophilin family with affinities ranging from ~5–500 nM (Davis et al., 2010). We observed dose-dependent lengthening of circadian period with low doses of CsA (Figures 5C, D), while at high concentrations (>15 μM), it induced arrhythmicity (data not shown). To probe the selectivity of CsA on cyclophilin regulation of the BMAL1 TAD, we performed two additional orthogonal assays. First, we examined whether the long period phenotype could be attributed to inhibition of the phosphatase PP2B, as CsA can also direct the assembly of inhibitory ternary complexes of cyclophilins with PP2B (Huai et al., 2002; Liu et al., 1991). The effect of CsA on circadian period appeared to be independent of PP2B, as we found that treatment with Deltamethrin, a cyclophilin-independent PP2B inhibitor (Enan and Matsumura, 1992) did not elicit a change in period (Figures 5C, D). Therefore, broad inhibition of isomerase activity in the cyclophilin family leads changes in circadian period.

We then wanted to determine the degree to which CsA-dependent changes in period arose from regulation of the BMAL1 TAD switch. To do this, we utilized stable Bmal1^−/−;Per2^Luc cell lines complemented with either wild-type Bmal1 or two trans-locked Bmal1 mutants, P625A or the W624A/P625A (Figure 2 and S5). We reasoned that any CsA-dependent period changes in the mutant lines could not be due to regulation of the switch by cyclophilins, as the cis isomer of the switch was eliminated upon introduction of these mutations (Figures 2 and S2). Consistent with this model, we observed a significant decrease in period lengthening by CsA in both trans-locked cell lines compared to cells complemented with wild-type Bmal1 (Figure 5E and S5). Switch-independent changes in period, particularly at the highest concentration of CsA tested (10 μM), demonstrate that regulation of other pathways aside from the TAD switch by cyclophilins can also influence circadian timing. Taken together, our data support a role for an intrinsically slow conformational switch in the BMAL1 TAD in regulation of circadian period and lay the foundation for studies of clock regulation by cyclophilins, a broadly expressed, yet poorly studied, class of enzymes.

DISCUSSION

Here we present our discovery of a slow conformational switch in the BMAL1 TAD that participates in regulation of timekeeping by the mammalian circadian clock. Using NMR spectroscopy, we identified that cis/trans isomerization about the W624-P625 imide bond at the C-terminus of BMAL1 underlies the molecular basis for this binary switch. Consistent with previous studies on the composition and rarity of other switches based on cis/trans isomerization (Shen and Bax, 2010; Stewart et al., 1990), we found that both proline and tryptophan side chains make critical contributions to the relatively high cis population observed in the BMAL1 TAD. These two residues are conserved from humans to invertebrates (including insects other than Drosophila) that have a vertebrate-like clock architecture (Chang et al., 2003), suggesting an ancient role for the TAD switch in regulation of CLOCK:BMAL1 activity. We used mutation of the Trp and Pro sites to generate a suite of TAD proteins with varying cis/trans populations or mutants that were altogether ‘locked’ into discrete cis or trans isomers. By coupling cell-based studies of circadian oscillations with solution biophysical techniques that probed structural and biochemical constraints of the TAD switch, we were able to demonstrate that perturbing switch function alters circadian timing. Moreover, we identified that cyclophilins can accelerate interconversion in vitro and influence circadian timing in clock cells in a switch-dependent manner, suggesting that cyclophilins regulate the clock at least partly through control of the TAD switch.

There is growing evidence that the flexible BMAL1 TAD plays an important role in control of circadian timekeeping. Given that transcription-based feedback loops establish the mammalian circadian clock, factors that influence how the CLOCK:BMAL1 interacts with regulators are likely to play a key role in timekeeping (Gustafson and Partch, 2015). The conformational switch we identified by NMR is located in the C-terminus of an intrinsically disordered transactivation domain that is essential for CLOCK:BMAL1 activity and circadian timekeeping (Park et al., 2015; Xu et al., 2015). Notably, the importance of this switch region was first highlighted a decade ago in a transposon-based screen to identify functionally important regions of BMAL1; Yagita and colleagues noted that truncation of the last 7 amino acids of BMAL1 decreased CLOCK:BMAL1 activity and impaired cycling (Kiyohara et al., 2006). We recently showed that this same region makes a significant contribution to binding both positive and negative transcriptional regulators by cooperating with a conserved alpha helical element in the TAD to influence clock timing (Xu et al., 2015). The data presented here identify yet another example of how the flexible BMAL1 TAD influences circadian timing and demonstrate the functional importance of protein dynamics at the TAD over an apparent twelve orders of magnitude.

Proline isomerization has demonstrated roles at each step of transcriptional regulation, from protein folding and regulation of transcription factor interactions, to modification and recognition of histone tails, and conformational control of the RNA Pol II C-terminal domain that controls its activity (reviewed in (Hanes, 2015)). Although we provide evidence in support of a binary conformational switch in the TAD by NMR and its ability to influence circadian timekeeping in cellular reconstitution assays, we still don't understand exactly how the switch regulates circadian period. Timekeeping by the circadian clock depends on the regulated transition through a series of transcriptional regulatory complexes on DNA throughout the day (Koike et al., 2012). This dynamic conformational switch could hinder the formation of highly stable complexes to play a role in the recognition and handoff between transcriptional coactivators and repressors at CLOCK:BMAL1. Other transcription factors such as p53, c-Myb and CREB utilize conformational changes at their TADs to control their activation state. While this most commonly arises from binding-induced changes in structure (Borcherds et al., 2014; Parker et al., 1999; Sugase et al., 2007), there is also evidence that slow events controlled by proline isomerization can play an important role in some cases (Follis et al., 2015). Our data demonstrate that both W624 and P625 establish the switch and contribute to normal circadian timekeeping. Interestingly, the W624A mutation not only locks the switch into its trans isomer, but also leads to an additional shortening of the period that is strikingly similar to the phenotype observed upon deletion of the entire switch region (Xu et al., 2015). Therefore, we believe that identifying the structural role that W624 plays in assembling complexes with transcriptional regulators will likely help identify how the TAD switch regulates CLOCK:BMAL1 activity.

To date, conformational control by proline isomerization has been relatively poorly studied due to the difficulty in identification and analysis of the process in vitro and in the cellular milieu. Similarly, our understanding of the peptidyl-prolyl isomerases that regulate this intrinsically slow process in the cellular context has lagged behind that of other signaling enzymes like kinases and phosphatases. We have a solid understanding how kinases, phosphatases and other enzymes that control chemical modification of clock proteins exert powerful roles in modulating clock timing (Gallego and Virshup, 2007). We provide evidence here that isomerases of the cyclophilin family accelerate the intrinsically slow cis/trans isomerization of the BMAL1 TAD by up to several hundred-fold in vitro. Moreover, studies with the broad specificity inhibitor CsA suggest that cyclophilins contribute to circadian timekeeping in cells in a TAD switch-dependent manner. Interestingly, we found that a number of cyclophilins that are active on the BMAL1 TAD are also expressed in vivo on a circadian timescale, suggesting the possibility for modes of feedback regulation that are common in circadian rhythms (Baggs et al., 2009). These findings are strengthened by the observation that animals with chronic administration of peptidyl-prolyl isomerases inhibitors exhibit defects in circadian cycling (Katz et al., 2008), as do organ transplant patients on long-term dosing of CsA (Kooman et al., 2001; van de Borne et al., 1993; van den Dorpel et al., 1996). More work is needed to parse out the potentially redundant roles of cyclophilins and possibly other proline isomerases on the BMAL1 TAD. However, our findings are consistent with two studies showing that the isomerase PPIE (also known as Cyp33) binds to the histone methyltransferase MLL1 to modulate its activity (Wang et al., 2010), which is in turn recruited to the CLOCK:BMAL1 complex in a circadian-dependent manner (Katada and Sassone-Corsi, 2010). Conceivably, MLL1-bound PPIE may also promote prolyl isomerization within the BMAL1 TAD when assembled within a CLOCK:BMAL1:MLL1 complex, lending credence to a model that these ubiquitous enzymes may be recruited to and act upon the molecular timer in a circadian fashion.

STAR Methods

CONTACT FOR REAGENTS AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Carrie Partch (cpartch@ucsc.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell lines

Isolated of the Bmal1^−/−; Per2^Luc line was previously described (Liu et al., 2008). The U2OS Bmal1-dLuc cell line was a gift from John Hogenesch (University of Cincinnati). Both cell lines were cultured in 10% DMEM (i.e. 10% (vol/vol) FBS) and 1X penicillin-streptomycin (Thermo Fisher) at 37°C in an incubator humidified with 5% CO₂. Cell lines were not authenticated.

METHOD DETAILS

Lentiviral DNA constructs, transduction, and analysis

Flag-tagged mouse Bmal1 was cloned into pENTR/D-TOPO vector (Life Technologies) then recombined with pLV7 destination vector as previously described (Ramanathan et al., 2012). Mutations were introduced by PCR-based mutagenesis and all constructs were verified by sequencing. Production of recombinant lentiviral particles, infection and selection of Bmal1-complemented Bmal1^−/−; Per2^Luc lines was performed as described previously (Liu et al., 2008). Clonal cell lines were validated by genomic sequencing of the complemented Bmal1 gene.

Expression of Flag-tagged Bmal1 genes in the complemented cell lines was analyzed as before (Xu et al., 2015). Briefly, cells were lysed in RIPA buffer containing complete protease and phosphatase inhibitors (Sigma). Immunoblotting was done using the following primary antibodies: mouse anti-Flag (M2) (Sigma Cat. # F3165) and goat anti-beta actin (C-11) (Santa Cruz Biotechnology Cat. # sc-1615), and the following secondary antibodies: anti-mouse IgG-HRP (Santa Cruz Biotechnology Cat. # sc-2005) and anti-goat IgG-HRP (Santa Cruz Biotechnology Cat. # sc-2020). SuperSignal West Pico substrate (Pierce) was used for chemiluminescent detection on autoradiograph film.

Bioluminescence recording and data analysis

For real-time recording of bioluminescence, Bmal1^−/−; Per2^Luc and U2OS Bmal1-dLuc cell lines were grown to confluence in 35 mm dishes in 10% DMEM at 37°C in an incubator humidified with 5% CO₂. All cell lines were synchronized with addition of 100 nM dexamethasone in recording medium, which contained phenol red-free DMEM with 25 mM HEPES, pH 7.4, 1% (vol/vol) FBS, 1 mM luciferin and a 1X B-27 vitamin supplement (Thermo Fisher Cat. #17504044). Dishes were sealed using vacuum grease with a round 40 mm glass coverslip to minimize evaporation, and then moved to a 37°C incubator (without humidification) containing a LumiCycle luminometer. For experiments with inhibitors, lyophilized stocks of Cyclosporin A (Sigma cat. # 30024) and Deltamethrin (Sigma cat. #D9315) were resuspended in sterile DMSO. Stocks were further diluted in DMSO such that addition of the same volume of inhibitors resulted in a final concentration of DMSO of 0.3% (vol/vol) in culture. Inhibitor stocks were pipetted into recording medium, evenly mixed, and then added to cells.

A LumiCycle luminometer (Actimetrics) was used to monitor the luminescence (counts/sec) as a function of time and data were analyzed using the LumiCycle Analysis program (version 2.54, Actimetrics). The first 24 hours of recording were omitted from data processing, and then the raw data were baseline corrected and fit to a damped sine wave, from which period length, goodness of fit, amplitude and damping rate were determined. Data were deemed acceptable if the goodness of fit exceeded 80%. For Bmal1^−/−; Per2^Luc cells, at least two dishes per clonal line were tested for each run with four repeats (n = 8–12). The mean trace of all recordings with standard deviation is reported.

Expression and purification of recombinant proteins

Mouse BMAL1 TAD (residues 579–626) was cloned into a bacterial expression plasmid based on the pET22b vector backbone from the parallel vector series (Sheffield et al., 1999). The BMAL1 short TAD (residues 594–626) was codon optimized for expression in E. coli (GeneWiz) and cloned into the same vector. All constructs possessed an N-terminal TEV cleavable His₆GST solubilizing tag and ampicillin resistance. Mutations were introduced in the TAD using a modified protocol for site-directed mutagenesis (Liu and Naismith, 2008) and confirmed with sequencing. The bacterial expression plasmid encoding the mouse CBP KIX domain (residues 585–672) was a gift from Peter Wright (The Scripps Research Institute). CBP KIX has native histidine residues that allow for the purification of the protein using Ni-NTA resin.

The Rosetta (DE3) strain of E. coli containing plasmids with either BMAL1 TAD or CBP KIX were grown to an OD₆₀₀ of ~0.6–0.9 in the presence of ampicillin (100 μg/mL) and chloramphenicol (35 μg/mL). Protein expression was induced with 0.5 mM isopropyl β-D-1-thiogalactopyranoside (IPTG) and allowed to proceed for 16–18 hours at 18°C in eithe r Luria Broth (LB) or M9 minimal medium containing 1g/L ¹⁵NH₄Cl to generate uniformly ¹⁵N-labeled proteins for NMR spectroscopy. Cells were lysed in with an Emulsiflex C-3 cell disruptor (Avestin) in Buffer A containing 50 mM Tris pH 7.5, 300 mM NaCl and 20 mM imidazole. The soluble fraction of E. coli lysates was passed over Ni-NTA resin (QIAGEN), washed thoroughly, and eluted using 250 mM imidazole. Fractions of interest were buffer exchanged into lysis buffer using a stirred-cell pressure concentrator with 3 kDa molecular weight cutoff filters (Amicon). Proteolysis was performed with His₆-tagged TEV protease overnight at 4°C and cleaved protein was retained from the flow-through of a Ni-NTA column. The purified protein was further purified on a preparative grade Superdex 75 16/600 size-exclusion column (GE Life Sciences) pre-equilibrated with NMR buffer (10 mM MES, pH 6.5 and 50 mM NaCl).

All of the bacterial expression constructs for human cyclophilins were previously described (Davis et al., 2010). Where possible, full-length proteins were expressed. However, isolated isomerase domains were expressed from the large, multidomain cyclophilins PPWD, PPIG, CWC27 due to solubility issues expressing full-length proteins. PPIH, PPIL3 and PPIG were cloned into the pET22b-based parallel vector system (Sheffield et al., 1999) with TEV-cleavable tags; PPIH and PPIL3 had a His₆GST tag to enhance yield and stability, and PPIG had a His₆ tag.

Rosetta (DE3) cells containing cyclophilin expression plasmids were grown at 37°C in LB with appropriate antibiotics until they reached OD₆₀₀ 0.8–1. Expression was induced with 0.5 mM IPTG and then the growth was continued overnight at 16°C. Cells were resuspended in Buffer A, lysed in a cell disruptor, and then purified by Ni-NTA column according to manufacturer’s protocol (QIAGEN). Tags were not cleaved with the exception of His₆GST-PPIH, which was subjected to overnight incubation with His₆-TEV protease at 4°C. The cleaved His 6GST tag was removed by passing the sample over Ni/NTA resin. All cyclophilins were further purified by running on a preparative grade Superdex 75 16/600 size-exclusion column pre-equilibrated with 20 mM HEPES pH 7.0, 100 mM NaCl, and 2 mM TCEP. For His₆GST-PPIL3, the fusion protein tag was not cleaved and the pH was increased to pH 7.5 to increase stability of the purified protein at room temperature.

Using the baculovirus expression system (Invitrogen), the His₆-tagged photolyase homology region (PHR, residues 1–491) of mouse CRY1 was expressed in Sf9 suspension insect cells (Expression Systems). Cells were infected with a high titer P3 virus at 1.5 × 10⁶ cells/mL and grown for 72 hours with gentle shaking at 27°C. Following brief centr ifugation at 4,000 rpm, cells were resuspended in 50 mM Tris pH 7.5, 200 mM NaCl, 20 mM imidazole, 10% (vol/vol) glycerol, 0.2% (vol/vol) Triton X-100, 0.1% (vol/vol) NP40, 0.4% (vol/vol) Tween-20, 5 mM β-mercaptoethanol and 1X EDTA-free protease inhibitors (Pierce). Cells were lysed using a cell disruptor followed by brief sonication on ice with a ¼” probe (3 pulses of 15 sec. on/30 sec. off). Lysate was clarified by centrifugation at 37,000 rpm at 4°C for 1 hour. His₆-CRY1 protein was isolated by Ni-NTA agarose affinity chromatography, and then further purified by size-exclusion chromatography on a preparative grade Superdex 200 16/600 column (GE Life Sciences) pre-equilibrated with 20 mM HEPES pH 7.5, 125 mM NaCl, 5% (vol/vol) glycerol and 2 mM TCEP. Prior to fluorescence anisotropy experiments, all purified proteins were buffer exchanged into assay buffer: 50 mM Bis-Tris Propane, 100 mM NaCl, 2 mM TCEP and 0.05% (vol/vol) Tween-20.

Peptide synthesis and purification

8-mer switch peptides DFSDLPWPL, FSDLPWPL, FSDLPAPL, FSDLPWAL were synthesized by solid phase peptide synthesis on 3-chlortrityl resin with standard Fmoc chemistry. One or two 1:4:4:4:6 molar ratio of resin:HBTU:HOAT:Fmoc-AA-OH:DiPEA coupling reactions were performed in DMF for each amino acid addition. The cis-locked peptide FSDLPWdmPL (dmP, 5,5-dimethyl L-proline) was synthesized by solid phase peptide synthesis using standard Fmoc chemistry. Coupling of dmP onto the Leu-Resin was performed using 1:2:2:4 molar ratio of Resin:HATU:Fmoc-dmP-OH:DiPEA, and coupling of the Trp onto the Resin-Leu-dmP was performed using 1:3.8:4:6 molar ratio of Rsin:COMU:Fmoc-Trp-Boc-OH:DiPEA; all other coupling reactions were performed using HBTU/HOAT as described above. Naturally occurring Fmoc-protected amino acids were purchased from Fluka, Nova Biochem, AAPPTec, or Sigma Aldrich. Fmoc-dmP was purchased from PolyPeptide Group (Cat. # FA21702). Peptides were purified by reverse phase C18 HPLC; purity (>90%) and identity were verified by MS/MS on a Waters HPLC-MS/MS system.

The BMAL1 short TAD (residues 594–626) containing the P625dmP mutation and 8-mer switch peptides FSDLPFPL, FSDLPYPL and the insect BMAL1 peptide FSGLPWPLP were purchased from Bio-Synthesis Inc. The mouse CRY1 CC peptide (residues 471–503) was synthesized and purified as described previously (Xu et al., 2015).

Mass spectrometry

The ¹⁵N BMAL1 TAD NMR sample was separated by a Surveyor HPLC system (Thermo Finnegan) with the Proto 300 C4 reverse-phase column (Higgins Analytical, Inc) with 5 μm particle size. A 20 μL aliquot of the sample was injected at flow rate of 200 μL/min with an autosampler tray at a temperature setting of 4°C. The mobile phase consisted of solv ent A: 0.1% formic acid in HPLC-grade water and Solvent B: 0.1% formic acid in acetonitrile with the following gradient: time (t) =0 to 3 min 95% solvent A and 5% solvent B; t = 28 min, 35% A and 65% B; t = 28.01 to 30 min, 5% A and 95% B; T= 30.01 to 40 min, 95% A and 5% B. The sample was then analyzed using a linear ion trap LTQ mass spectrometer system (Thermo Finnegan). Proteins were detected by full scan MS mode (over the m/z 300–3000) in positive mode. The electrospray voltage was set to 5 kV. Mass measurements of deconvoluted ESI mass spectra of the reversed-phase peaks were generated by Magtran software (Zhang and Marshall, 1998).

NMR spectroscopy

NMR experiments were conducted on a Varian INOVA 600-MHz spectrometer equipped with ¹H, ¹³C, ¹⁵N triple resonance, Z-axis pulsed field gradient cryoprobe. Sample temperatures were calibrated with the use of an ethylene glycol standard supplied by Agilent. At each temperature, NMR samples were given a 30-minute equilibration time prior to calibration and data acquisition. All NMR data were processed using NMRPipe/NMRDraw (Delaglio et al., 1995). Chemical shift assignment of mBMAL1 TAD was previously reported (Xu et al., 2015). ¹⁵N-¹H HSQC titrations of BMAL1 TAD were performed in a 300 μL volume of 100 μM ¹⁵N BMAL1 TAD in NMR buffer (10 mM MES pH 6.5, 50 mM NaCl) with 10% (vol/vol) D₂O by stepwise addition of CBP KIX or CRY1 CC peptide, followed by concentration in an Amicon Ultra centrifugal filter with a 3 kDa molecular weight cutoff. All titration data were collected at 25°C. Titration data were analyzed with NMRViewJ (One Moon Scientific) using chemical shift perturbations defined by the equation Δδ_TOT = [(Δδ¹H)² + (χ(Δδ¹⁵N)²]^½ and normalized with the scaling factor χ = 0.5 (Johnson, 2004).

¹⁵N-¹H ZZ-exchange experiments (Farrow et al., 1994) were collected on 400 μM ¹⁵N BMAL1 TAD in NMR buffer (10 mM MES pH 6.5, 50 mM NaCl) with 10% (vol/vol) D₂O with 24 interleaved mixing times ranging from 0–3 seconds (0, 0.05, 0.1, 0.25, 0.5, 0.75, 1,1.25, 1,5, 1.75, 2, 2.5, 3) at temperatures of 25, 35, 45, 55, 60 and 65°C. Integration of the auto and cross peak intensities for each mixing time were extracted using NMRViewJ (One Moon Scientific). Cross peak intensities were normalized to the nominal cross peak intensities at time t=0, and total intensity was set to the sum of the integrations of the cis and trans peaks at t=0. The exchange constant k_ex was calculated by fitting integration data to an exchange model for two-state interconversion as described by equations 1–4 (Kleckner and Foster, 2011; Palmer et al., 2001) using MATLAB (MathWorks).

$$I_{\text{AA}}(T)=\frac{1}{2}{P}_{\mathrm{A}}\lfloor \left(\frac{\left(1-{{\mathrm{R}}^0}_{1\mathrm{A}}-{{\mathrm{R}}^0}_{1\mathrm{B}}+k_{\text{ex}}({\mathrm{\rho }}_{\mathrm{B}}-{\mathrm{\rho }}_{\mathrm{A}}\right))}{{\lambda }_{+}-{\lambda }_{-}}\right)e^{-{\mathrm{\lambda }}_{-}t}+(1+\left(\frac{\left(1-{{\mathrm{R}}^0}_{1\mathrm{A}}-{{\mathrm{R}}^0}_{1\mathrm{B}}+k_{\text{ex}}({\mathrm{\rho }}_{\mathrm{B}}-{\mathrm{\rho }}_{\mathrm{A}}\right))}{{\lambda }_{+}-{\lambda }_{-}}\right)e^{-{\mathrm{\lambda }}_{=}t}\rfloor$$ (1)

$$I_{\text{BB}}(T)=\frac{1}{2}{P}_{\mathrm{B}}\lfloor \left(\frac{\left(1-{{\mathrm{R}}^0}_{1\mathrm{A}}-{{\mathrm{R}}^0}_{1\mathrm{B}}+k_{\text{ex}}({\mathrm{\rho }}_{\mathrm{B}}-{\mathrm{\rho }}_{\mathrm{A}}\right))}{{\lambda }_{+}-{\lambda }_{-}}\right)e^{-{\mathrm{\lambda }}_{-}t}+(1+\left(\frac{\left(1-{{\mathrm{R}}^0}_{1\mathrm{A}}-{{\mathrm{R}}^0}_{1\mathrm{B}}+k_{\text{ex}}({\mathrm{\rho }}_{\mathrm{B}}-{\mathrm{\rho }}_{\mathrm{A}}\right))}{{\lambda }_{+}-{\lambda }_{-}}\right)e^{-{\mathrm{\lambda }}_{=}t}\rfloor$$ (2)

$$I_{\text{AB}}(T)=P_{\mathrm{B}}\lfloor \left(\frac{\left(k_{\text{ex}}{P}_{\mathrm{A}}\right)}{{\lambda }_{+}-{\lambda }_{-}}\right)\left(e^{-{\mathrm{\lambda }}_{-}t}-e^{-\lambda =t}\right)\rfloor$$ (3)

$$I_{\text{BA}}(T)=P_{\text{BA}}\lfloor \left(\frac{\left(k_{\text{ex}}{P}_{\mathrm{B}}\right)}{{\lambda }_{+}-{\lambda }_{-}}\right)\left(e^{-{\mathrm{\lambda }}_{-}t}-e^{-{\mathrm{\lambda }}_{=}t}\right)\rfloor$$ (4)

I refers to the time dependence of the transfer amplitudes (represented by the build-up curves) for the cis (AA), trans (BB) and trans to cis (BA), and cis to trans (AB) interconversions. P refers to the population of the indicated state, k_ex is the stochastic exchange of molecules between the two states per second, t is time in seconds, T is temperature in Kelvin, and R1A and R1B are the longitudinal relaxation rate constants in the absence of exchange.

The interconversion rates of cis to trans (k₋₁) and trans to cis (k₁) were calculated using the relative populations of the two isomers taken from the integrations using equations 5–7.

$$k_{\text{ex}}=k_1+k_{-1}$$ (5)

$$k_1=k_{\text{ex}}\ast P_{\text{trans}}$$ (6)

$$k_{-1}=k_{\text{ex}}\ast P_{\text{cis}}$$ (7)

Rates of isomerization were extrapolated to 25°C and 37°C using the Eyring equation (equation 8). The free energy of isomerization was calculated based on transition state theory using equation 9 and the difference in free energy between the two isomers calculated using equation 10.

$$ln\frac{K}{T}=-\frac{\Delta H}{RT}+ln\frac{kB}{h}+\frac{\Delta S}{R}$$ (8)

$$\Delta G^{\ne }=-\mathit{RTln}\left(\frac{h{K}_{\text{CT}}}{k_{\mathrm{B}}T}\right)$$ (9)

$$\Delta G=|\Delta {G^{\ne }}_{\text{CT}}-\Delta {G^{\ne }}_{\text{TC}}|$$ (10)

T is temperature, R is the gas constant, k_B is the Boltzmann constant, h is Planck’s constant, and ΔH and ΔS are the activation enthalpy and entropy, respectively, of cis to trans (CT) and trans to cis (TC) isomerization. The free energy of the isomerization was calculated using equation 8.

To assess rate enhancement of interconversion by cyclophilins, ¹⁵N-¹H ZZ-exchange data were collected on 400 μM ¹⁵N BMAL1 TAD with 100 μM cyclophilin (natural abundance) in cyclophilin NMR buffer, 20 mM HEPES pH 7.0, 100 mM NaCl, and 2 mM TCEP with 10% (vol/vol) D₂O. As described above, the pH of the buffer was increased to pH 7.5 for His₆GST-PPIL3 to increase stability of the purified protein at room temperature. ZZ data were collected as above, except that experiments were performed at room temperature (22.68°C). The fold enhancement of isomerization rates was determined by comparing the uncatalyzed rates with catalyzed rates extrapolated to room temperature for the isolated BMAL1 TAD.

Fluorescence anisotropy

The BMAL1 short TAD WT, P625A, P625dmP and Δswitch (594-F619Y) probes were purchased from Bio-Synthesis Inc. with a 5,6-TAMRA fluorescent probe covalently attached to the N-terminus. The C-terminus of the Δswitch peptide was amidated, while the others were left as a free carboxyl group to mimic the native C-terminal group of the TAD at L626. Equilibrium binding assays with CRY1 PHR were performed in 50 mM Bis-Tris Propane pH 7.5, 100 mM NaCl, 2 mM TCEP and 0.05% (vol/vol) Tween-20. With CBP KIX, the assay was performed in 10 mM MES, pH 6.5, 50 mM NaCl and 0.05% (vol/vol) Tween-20. Concentrated stocks of BMAL1 TAD probes were stored between 15–200 μM at −70°C and diluted into assay buffer to 50 nM alone and in the presence of increasing concentrations of test proteins. Plates were incubated at room temperature for 10–20 minutes prior to analysis. Binding was monitored by changes in fluorescence polarization with an EnVision 2103 multilabel plate reader (Perkin Elmer) with excitation at 531 nm and emission at 595 nm. The Hill coefficient (n_H) and equilibrium binding dissociation constant (K_D) were calculated by fitting the dose-dependent change in millipolarization (Δmp) to a one-site specific binding model in Prism 6.0 (GraphPad), with averaged Δmp values from duplicate or triplicate assays. Data shown are from a representative experiment (n=4 replicates) of three independent assays.

Isothermal titration calorimetry

ITC measurements were obtained as previously described (Xu et al., 2015). Briefly, proteins were extensively dialyzed at 4°C in 10 mM MES pH 6.5, 50 mM NaCl using a 2 kDa molecular cutoff filter dialysis tubing (Spectrum Labs) prior to collecting ITC data. ITC was performed on a MicroCal VP-ITC calorimeter at 25°C with a stir speed of 177 r.p.m., reference power of 10 μCal/sec and 10 μL injection sizes. Protein ratios for the cell and syringe for the ITC assays (3 independent runs for each complex) were 220–230 μM CBP KIX titrated into 20–25 μM BMAL1 TAD WT, P625A, or P625dmP (N = 0.6–0.9). All data were best fit by a one-site binding model using Origin software.

QUANTIFICATION AND STATISTICAL ANALYSIS

Where applicable, statistical parameters including sample size, precision measures (standard deviation, s.d.) and statistical significance are reported in the Figures and corresponding Figure Legends.

Cellular Period Analysis

A LumiCycle luminometer (Actimetrics) was used to monitor luminescence (in counts/sec) as a function of time, and data were analyzed using the LumiCycle Analysis program (version 2.54, Actimetrics). The first 24 hours of recording were omitted from data processing, and then the raw data were baseline corrected and fit to a damped sine wave, from which period length, goodness of fit, amplitude and damping rate were determined. Data were deemed acceptable if the goodness of fit exceeded 80%. For Bmal1^−/−; Per2^Luc cells, at least two dishes per clonal line were tested for each run with four repeats (n = 8–12). The mean trace of all recordings with s.d. is reported. Statistical significance between genotypes or drug treatments was assessed using an unpaired t test in Prism 6.0.

Analysis of cis/trans isomerization rates

Experimentally determined rate constants for isomerization were visualized on an Eyring plot with representative s.d. errors from the derivation of individual cis-trans and trans-cis exchange rates from the time series NMR data. Data points at 55°C, 60°C, and 65°C were fitted to a linear regression to extrapolate exchange rates, not fast enough to detect experimentally by NMR, at lower temperatures (e.g. 25°C and 37°C).

DATA AND SOFTWARE AVAILABILITY

Raw data from the following ¹⁵N-¹H ZZ exchange NMR time series experiments on ¹⁵N BMAL1 TAD is available at Mendeley Data:

• 25°C time series: doi:10.17632/4hs472w6wb.1

• 55°C time series: doi:10.17632/8yx9wzt9gh.1

• 60°C time series: doi:10.17632/gjfjvjpvdr.1

• 65°C time series: doi:10.17632/kn58vc7hww.1

• room temperature time series with PPIA: doi:10.17632/4txk4b3w5g.1

• room temperature time series with PPIE: doi:10.17632/pt8xjvjs2v.1

• room temperature time series with PPIF: doi:10.17632/x8yn6x28x8.1

• room temperature time series with PPIG: doi:10.17632/fh595ry88h.1

• room temperature time series with PPIH: doi:10.17632/mpykzm4kw2.1

• room temperature time series with PPIL1: doi:10.17632/f9fxfynfrb.1

• room temperature time series with PPIL2: doi:10.17632/cg8fb64p5f.1

• room temperature time series with PPIL3: doi:10.17632/fmhg69tcxf.1

• room temperature time series with PPWD: doi:10.17632/csc3c5hs32.1

• room temperature time series with CWC27: doi:10.17632/htvhkdt7rw.1

An algorithm for analysis of ZZ exchange NMR time series data (and instructions for its use in MATLAB) are available at Mendeley Data: doi:10.17632/w7ytgp2y9d.1.

Table 2.

Kinetics of cis/trans isomerization at the TAD switch as measured by ¹⁵N-¹H ZZ-exchange NMR spectroscopy

	Rate of isomerization (s−1)		Time per isomerization event (s)
Temperature (°C)	cis➔trans	trans➔cis	cis➔trans	trans➔cis
25*	4.58 × 10−3	2.64 × 10−3	218.34	378.78
37*	2.33 × 10−2	1.35 × 10−2	42.92	74.07
55	2.15 × 10−1	1.24 × 10−1	4.65	8.06
60	3.82 × 10−1	2.20 × 10−1	2.62	4.55
65	6.67 × 10−1	3.84 × 10−1	1.50	2.60

^*Values extrapolated from Eyring plot