An evolutionary hotspot defines functional differences between CRYPTOCHROMES

To investigate the possibility that functional coupling between the primary and secondary pockets is a conserved feature across both the CRYS and PHLs, we conducted the statistical coupling analysis (SCA), a method for inferring groups of conserved, co-evolving positions associated with complex functions such as allostery in a protein family^40,41. These co-evolving groups are called sectors, and typically comprise physically contiguous networks of amino acids that link active sites to distantly positioned surfaces. Sectors correspond to allosteric mechanisms in several model systems, and recent work has shown that sectors create hotspots for allosteric regulation, enabling the evolution of new regulation in proteins through local modifications^42,43. While distinct co-evolution-based approaches have also been developed to identify direct structural contacts in proteins (e.g., DCA, PSICOV, and GREMLIN), SCA uniquely focuses on discovering the collective networks of residues that underlie long-range coupling within a protein.

A key feature of SCA is the ability to mathematically relate the amino acid motif within sectors (correlated evolution of positions) to phylogenetic and/or functional divergences in the protein family (correlated evolution of sequences)^44,45. This relationship permits us to test whether an allosteric mechanism inferred by SCA is conserved or is divergent in members of the protein family. For example, the family of Hsp70-like molecular chaperones contains both allosteric and non-allosteric members, and sector positions display the property of (1) cleanly separating these functional subfamilies and (2) displaying lower conservation in the non-allosteric group⁴⁵. In contrast, the finding that sector positions do not separate subfamilies by function and do not show differences in overall conservation in subfamilies would be consistent with retention of an allosteric mechanism or a conserved functional coupling between sites. For the CPF, we expect SCA to reveal a sector linking the FAD-binding pocket (or primary pocket) with the secondary pocket. If so, we can then use the sector positions to test the functional role of the secondary pocket in the CRY subfamily.

SCA for a multiple sequence alignment of 5385 CPF proteins reveals a single sector of 260 co-evolving amino acid positions that surrounds the primary FAD-binding site and extends to several surface accessible positions, including the secondary cofactor pocket (Fig. 1d and Supplementary Table 1). These findings are consistent with the known functional coupling between these sites in the PHLs. Interestingly, sequence logos characterizing the distribution of amino acids in sector positions comprising the primary and secondary pockets in PHLs and CRYs show similar patterns of positional conservation (Fig. 1e, f). This finding provides one level of support for the maintenance of functional coupling between the two pockets in these functional subfamilies.

We mapped the network of residues identified by SCA onto structures of the CRY2/PER2 CRY-binding domain (CBD) complex (Supplementary Fig. 1a) and the CRY2/FBXL3 complex (Supplementary Fig. 1b)^20,24. Consistent with repurposing of the PHL active site in the CRYs, we found that PER2 and FBXL3 interact extensively with sector positions near the FAD-binding pocket (where DNA lesions directly interact with PHLs) and around CRY2’s CC helix. A logical hypothesis is that the secondary pocket has also been repurposed for regulatory interactions with CLOCK/BMAL1, with the residues in the yellow subgroup representing candidates for specifically contributing to the allosteric mechanism.

In summary, SCA exposes a co-evolving network of amino acids consistent with the known mechanism in the PHLs, and provides evidence for the conservation of functional coupling between the primary and secondary pockets in the CRY proteins as well. Since SCA provides residue-level resolution, we also have a logical basis for interrogating specific amino acid positions that are predicted to contribute to the allosteric mechanism.

In contrast, even severe mutations of the non-sector residues at positions 107 (K107E) and 108 (E108K) of CRY1 did not impair CRY1’s ability to rescue rhythms (Fig. 3e, f), and these mutants interacted strongly with CLOCK, BMAL1, and PER2 (Fig. 3d and Supplementary Figs. 2 and 10a), suggesting that these residues are not critical for binding to CLOCK and BMAL1. Consistent with this interpretation, K107 and E108 are superficial residues with side chains that neither gate access nor form any part of the surface of the interior of the pocket. Based on these results, we suggest that the interior of the secondary pocket is a critical interface for interactions between CRYs and their repressive targets CLOCK and BMAL1, consistent with another recent report¹⁹. Ultimately, in concert with this report, we conclude that the secondary pocket is a site for interaction with CLOCK specifically. These results also support the predictions from SCA that a subset of residues in the secondary pocket primarily mediate regulation of this protein–protein interaction.

We tested the hypothesis that differences at these two interfaces partially underlie the more direct role of CRY in vertebrate clocks. Single amino acid substitutions in CRY1 of the divergent residues at positions identified in the SCA (P39G, F41S, G106W) caused either weakly repressive, short period rescues (P39G), or arrhythmicity (F41S, G106W) (Fig. 4c, d). Substitution at a convergent residue (D38A) from the SCA network also resulted in no rescue (Fig. 4c). All of these substitutions resulted in attenuated interactions with CLOCK and BMAL1, though P39G retains the strongest interaction (Fig. 4e and Supplementary Fig. 4). Interactions with PER2 were mostly intact, though there was a trend toward a weakened interaction between PER2 and the F41S and G106W substitutions (Fig. 4e and Supplementary Figs. 4 and 10b). Taken together, these data suggest that the secondary pocket of type II CRYs has evolved a more accessible conformation through which CRY interacts with the CLOCK/BMAL1 heterodimer. This interaction defines a critical feature of vertebrate-like clocks and likely underlies, at least in part, the direct role of CRY in vertebrate-like clocks.

We hypothesized that these period-shortening pocket mutations modulate the affinity between the CLOCK/BMAL1 heterodimer and CRY1. Though there was a trend towards weaker interaction, co-IP of CLOCK and BMAL1 in the presence of overexpressed PER2 was not significantly affected (Fig. 5g and Supplementary Figs. 5c and 10c) and none of the mutants showed significant changes in PER2 binding when overexpressed only with PER2 (Fig. 5h and Supplementary Figs. 5d and 10d). However, when overexpressed with just CLOCK and BMAL1, some of these mutants showed a significant decrease in interaction with CLOCK and BMAL1 in proportion to overall shortening of the rescue period (Fig. 5i and Supplementary Figs. 5e and 10e). These results imply that mutations in the secondary pocket can modulate the affinity between CRY1 and the CLOCK/BMAL1 heterodimer, which in turn tunes the period of the oscillation. This effect is most obvious without PER2, suggesting that the shortened period in these mutants might stem from an increasing requirement of PER co-expression to stabilize the repressive interaction of CRY with CLOCK and BMAL1.

In summary, these results argue that through regulation at a distance, variations in the secondary pocket can provide a fine-tuning of CRY function through modulation of CLOCK/BMAL1 binding.

Given our finding that substitutions in this region of CRY1 can cause substantial shortening of the period of the circadian oscillation, we hypothesized that several subtle changes at this interface might underlie the shorter period in clocks driven only by CRY2. Converting any of these five residues in CRY1 individually to its CRY2 homolog had a minimal effect on the periodicity of the rescue (Supplementary Fig. 6a, b) with the greatest effect observed in E376S and W390F rescues, which were accelerated by roughly an hour per day (Supplementary Fig. 6b). However, combinatorial conversion of any of these residues to their CRY2 homologs led to increasingly short rescue periods, including the three residues that had no effect at the individual level (M378V/K379R/E382D, i.e., CRY1 3 m) (Fig. 6b, c). Mutating all five residues in concert (CRY1 5m) led to the most substantial shortening compared to WT (24.4 h vs. 26.9 h) (Fig. 6c). Examination of the residues contributing to the surface area of the pocket shows that in addition to the five residues identified above, there are only two other divergent residues: G43/A61 and N46/S64 (Fig. 6a).

Mutation of both residues also led to an accelerated rescue rhythm compared to WT CRY1 (Supplementary Fig.). Furthermore, combinatorial mutation of these residues also resulted in a corresponding weaker repression of CLOCK/BMAL1 transcriptional activation as demonstrated by the higher overall luminescence in these rescues (Supplementary Fig.). 6c, d 6c

Khan et al.²⁷ demonstrated that even under the control of the Cry1 promoter, Cry2 was unable to rescue rhythms in Cry-deficient cells. We reasoned that if the seven residues at this pocket differentiate CRY1 and CRY2 at a functional level, then mutation of all seven together in CRY1 should cause it to fail to rescue rhythms. As a control for this experiment, full-length, WT Cry2 was cloned into the Cry1 rescue vector. However, we found that both the seven-residue pocket mutant (CRY1 7m) and CRY2 were competent to rescue rhythms (Fig. 6d). Moreover, both CRY1 7m and CRY2 produced rhythms with significantly shortened periods (23.48 h and 22.97 h, respectively) compared to WT CRY1 (27.25 h) (Fig. 6e), consistent with in vivo data from Cry1^−/− mice, which also have short periods of ~22.5 h^8,9. Compared to WT CRY1, both CRY2 and CRY1 7m were substantially derepressed and had higher amplitude oscillations. These results suggest that subtle divergence at this interface is a major driver of period differences in CRY1-driven or CRY2-driven rhythms. To test this hypothesis further, all seven residues in CRY2 were converted to their CRY1 homologs (denoted CRY2 7m). The effect of these mutations was again additive (Supplementary Fig. 6e, f), and resulted in a significant lengthening of the period (24.22 h) and an increase in repression compared to WT CRY2 (Fig. 6d, e), though these effects were more modest than the CRY1 7m mutant.

Given that we were using the same cell line as Khan et al.²⁷ and a nearly identical Cry2 rescue construct, we were surprised to see a strong rescue by CRY2. Despite testing the few obvious variables that we could identify between these experiments, we ultimately could not clearly determine the reason underlying diverging reports of Cry2’s capacity to rescue rhythms^22,27,50.

To determine whether the pocket architecture affects the intrinsic stability of CRY1 and CRY2, we measured the half-life of CRY1, CRY2, CRY1 7m, and CRY2 7m, and found that, consistent with a previous report, CRY2 is more stable than CRY1⁵⁰ (Fig. 6f, g). However, neither CRY1 7m nor CRY2 7m differed significantly from CRY1 or CRY2, respectively, in half-life, suggesting that the period differences seen in the rescue assays were not due to changes in stability. Moreover, we found no correlation between half-life and period for any of the mutant residues in the secondary pocket that rescue rhythms (Fig. 6h). We also performed rescues with Cry1, Cry2, Cry1 7m, and Cry2 7m over a broad range of doses (25 ng to 1200 ng of rescue vector) and found a small effect of dose on period length (Supplementary Fig. 7). However, the results consistently show a difference between Cry1 and the other rescues at all doses, suggesting that differences in DNA dosage are not driving the divergent periods.

Several recent reports characterizing the interaction between CRY and CLOCK¹⁹ and CRY and BMAL1¹⁸ suggest that the secondary pocket is gating the interaction between the heterodimer primarily through a direct interaction with CLOCK. Taking into account these results along with our data suggesting that timing differences between CRY1 and CRY2 are primarily generated by divergence at the secondary pocket, we chose to focus on CLOCK rather than BMAL1 to test whether CRY1 has a stronger interaction than CRY2 with CLOCK/BMAL1 in a cellular milieu. We performed a reciprocal two-color bimolecular fluorescence complementation (BiFC) competition assay (Fig. 7d). In this assay, a C-terminal Cerulean (CerC) fragment can interact with either an N-terminal Venus (VenN) or N-terminal Cerulean (CerN) fragment to produce fluorescence in the Venus or Cerulean range, respectively. Non-complementary controls were used to establish background fluorescence as previously described¹⁵. A CerC-CLOCK construct containing the PAS-A and PAS-B domains was expressed with VenN-tagged and CerN-tagged CRYs and we found that CRY1, regardless of its tag, interacted with CLOCK more strongly than CRY2 (Fig. 7e, f), which suggests that the interactions seen in our co-IPs are relevant in vivo. These data strongly support the idea that the seven unique residues at the secondary pocket of CRY1 and CRY2 are critical for gating a strong physical interaction with CLOCK and BMAL1. They also suggest that expression of PER restricts the phase of repression for CRY2 by facilitating the formation of a stable repressive complex, while CRY1 is able to maintain a repressive complex even as PER levels decrease.

Primers used in this study are detailed in Supplementary Table. Key resources used in this study are shown in Supplementary Tablealong with manufacturer details where relevant. 2 3

Cry1^−/−/Cry2^−/− mouse embryonic fibroblasts were generated by Andrew Liu and Hiroki Ueda⁴⁶ and were gifted to us. HEK-293A cells were purchased from Thermo Fisher Scientific. Both lines were grown in Dulbecco’s modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS) and 1× Pen/Strep antibiotics at 37 °C under 5% CO₂. LumiCycle recording medium was prepared from powdered DMEM without phenol red containing 4.5 g/L glucose and supplemented with 10 mM HEPES, pH 7.20, 100 μM luciferin, 1 mM sodium pyruvate, 0.035% sodium bicarbonate, 2% FBS, 1× Pen/Strep antibiotics, and 2 mM l-glutamine. LumiCycle recordings were performed at 37 °C.

Approximately 10 000 CRY and PHL sequences were collected from NCBI on March 30, 2016 using the full-length mCRY1 protein sequence (GenBank Accession ID: AAH85499.1↗) as a search sequence for a PSI-BLAST. PSI-BLAST parameters were adjusted from default to include an Expect Threshold of 0.01, a PSI-BLAST Threshold of 0.005, and a minimum sequence identity of 20%. PSI-BLAST was performed iteratively for two rounds before the sequences were downloaded. The sequences were initially filtered by size using a custom python script to remove sequences smaller than 400 or larger than 800 residues. Remaining sequences were subjected to an initial alignment using the alignment tool MUSCLE⁶⁹ locally for two iterations. Extraneous header information was removed using a custom python script before performing a secondary alignment using Promals3D with six additional structure sequences (cyclobutane pyrimidine dimer (CPD) PHL from Escherichiacoli, CPD PHL from Anacystisnidulans, 6–4 PHL from Arabidopsisthaliana, CRY from Drosophilamelanogaster, and CRY1 and CRY2 from Musmusculus^{24,26,51,70–73}. Using a custom python script, GI numbers were collected from each header in a separate file for use in the annotation step of the SCA. SCA calculations were performed using the pySCA toolbox as described⁴⁴. The annotation step was performed through NCBI using the GI numbers on April 17, 2016. The initial alignment of 9719 sequences consisted of 4344 independent positions. The alignment underwent preprocessing during which highly gapped positions were removed. After removing highly gapped positions, remaining highly gapped sequences were removed using a set cutoff. Sequences with too great of a fractional identity to the reference sequence were also removed. Finally, each sequence was weighted based on the number of sequences with an identity above 80% to the given sequence. This final weighting step allows an effective number of sequences (M′) to be computed based on the remaining number of actual sequences (M). After all of the preprocessing steps, the final alignment contained 5385 sequences representing 2447 effective sequences composed of 459 positions. Critically, an alignment should be large and diverse enough to give reasonable estimates of amino acid frequencies. Two thousand four hundred and forty-seven effective sequences is well above the suggested minimum of 100 effective sequences needed to obtain a reasonable estimate⁴⁴. During these processing steps, default parameters were used and residues were mapped to the CRY2 structure (PDB: 4I6E)²⁴. Following initial calculation steps, the workflow was performed in an ipython notebook.

Here we provide a brief overview of the SCA calculations; the complete mathematical formalism and methods as used here are more completely presented in ref ⁴⁴. Conservation values for individual amino acid positions were calculated as a Kullback–Leibler relative entropy (D_i), a measure that quantifies the divergence of amino acid frequencies at position i from the distribution of amino acid frequencies expected randomly. Values of D_i near zero indicate that amino acid frequencies at that site approach random expectation, while values of more than 3 indicate highly conserved positions (>~83% residue identity). Co-evolution between all pairwise combinations of amino acids over all L positions was computed as an L x L x 20 × 20 conservation-weighted co-variance matrix:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde {C_{ij}^{ab}} = \frac{{\partial D_i^a}}{{\partial f_i^a}}\frac{{\partial D_j^b}}{{\partial f_j^b}}\left( {f_{ij}^{ab} - f_i^af_j^b} \right),$$\end{document}Cijab ~=∂Dia∂fia∂Djb∂fjbfijab-fiafjb,where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$f_i^a$$\end{document}fia indicates the frequency of amino acid a at site i (see also ref. ⁴⁴). The basic idea is that positions experiencing a joint functional constraint will co-vary in an alignment; weighting the co-variance by the degree of conservation \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\left( {{\partial D_i^a}}/{{\partial f_i^a}} \right)$$\end{document}∂Dia∕∂fia serves to emphasize couplings between functionally relevant positions and helps reduce phylogenetic noise. This matrix was then dimension reduced to produce the final SCA matrix:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde {C_{ij}} = \sqrt {\mathop {\sum}\limits_{a,b} {\left( {\widetilde {C_{ij}^{ab}}} \right)^2} },$$\end{document}Cij ~= ∑a,bCijab ~2,

In order to identify correlated groups of positions within the protein, we analyzed the SCA matrix \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde {C_{ij}}$$\end{document}Cij ~ using spectral decomposition. We defined a single sector along the top eigenmode (EV1 > 0.03) as shown in Figs. 1 and 2 and Supplementary Table 1. The top two eigenvectors of the positional correlation matrix were then used to project the correlations between sequences to arrive at Fig. 2b, c.

Mutagenesis was performed using both the QuikChange II XL kit and, with a few modifications, the Q5 Site-Directed Mutagenesis kit. Manufacturer's instructions were followed for QuikChange mutagenesis. For Q5 Site-Directed Mutagenesis, primers were designed using the NEBaseChanger tool (http://nebasechanger.neb.com/↗). Template DNA was mixed with 2× Q5 polymerase chain reaction (PCR) Master Mix and primers before being subjected to PCR. Following PCR, 1 μL of each PCR product was combined with 2× Quick Ligation buffer and ligase, along with 1 μL of T4 polynucleotide kinase, and 1 μL of DpnI in a total of 10 μL. The ligation reaction was incubated at room temperature for 5 min and 5 μL of product were used to transform DH5α-competent cells. Colonies were selected for culturing and miniprep, and mutations were verified by Sanger sequencing.

CRY tail chimera rescue vectors were generated by Gibson assembly cloning using NEBuilder HiFi DNA Assembly kit. Briefly, the rescue vector containing Cry1, Cry2, Cry1 7m, or Cry2 7m was linearized by PCR with primers, which also removed the coding sequence for residues 498–606 of Cry1 and Cry1 7m or residues 516–593 of Cry2 and Cry2 7m. The C-terminal Myc tag was left intact. The coding sequences for the tail regions of Cry1 (residues 498–606) and Cry2 (residues 516–593) were amplified by PCR using primers that would generate overlaps between the amplified tail product and the linearized target vectors. PCR products were purified using the QIAquick PCR Purification kit and combined with inserts and linearized target vectors in a 2:1 molar ratio. These combined products were then treated with NEBuilder HiFi DNA Assembly master mix containing an exonuclease, DNA polymerase, and DNA ligase to induce assembly of the final vector. This solution was incubated at 50 °C for 15 min and the products were used to transform DH5α competent cells. Colonies were selected for culturing and miniprep, and insertions were verified by Sanger sequencing.

A total of 4 × 10⁵Cry1^−/−/Cry2^−/− mouse embryonic fibroblasts were plated in 35 mm tissue culture dishes and transfected the same day with 4 μg of a luciferase reporter (pGL3-P(Per2)-Luc) and 150 ng of a Cryptochrome rescue vector (pMU2-P(Cry1)-(intron336)-Cry-Myc, modified with a C-terminal MYC tag)⁴⁶ using FuGENE 6 according to the manufacturer's instructions. At 72 h after transfection, the cells were synchronized by exchanging growth medium for growth medium supplemented with 0.1 μM dexamethasone and returned to the incubator for 2 h. The medium was then replaced by LumiCycle recording medium and the plates were sealed with vacuum grease and cover glass and transferred to the LumiCycle. Bioluminescence monitoring was performed using a LumiCycle to record from each dish continuously for 70 s every 10 min using a photomultiplier tube at 37 °C. Rescue results were processed using the LumiCycle Analysis software package. The first 10 h of recording were discarded and period, amplitude, phase, and damping rate were calculated using a damped sine wave based on a running average fit for each plate of cells. Rescues were considered arrhythmic if the goodness of fit for the damped sine wave was <80%. Subsequent data were corrected for background noise in each photomultiplier tube channel, which was measured by monitoring an untransfected plate of cells for 24 h and averaging the signal. All CRY1 and CRY2 mutants were run with WT CRY1 as an internal control. In some cases, multiple experiments were combined in the data shown if there were positive internal controls for each dataset. Rescues were excluded only in cases where the transfection did not work, which was obvious based on both a very weak or non-existent luminescence signal and the lack of induction of luminescence shortly after synchronization. Appropriate replicate size was determined empirically from extensive use of these assays in previous work²⁰.

Full-length mCry2 cDNA was cloned into the pMU2-P(Cry1)-(intron336)-Myc rescue vector backbone using megaprimer mutagenesis. Luciferase-tagged constructs were also generated using megaprimer mutagenesis. Briefly, the relevant insert was copied from a source vector by PCR (using PfuUltra II Fusion HS DNA polymerase with primers containing roughly 30 bp of overlap in either direction with the insertion site in the target vector and 30 bp of overlap with the insert). Following primary PCR, PCR products were purified using the QIAquick kit, and 400 ng of purified product used as a megaprimer for a secondary PCR with the target vector serving as a template. PCR products were treated with DpnI for several hours, and then DH5α-competent cells were transformed with 5 μL. Colonies were selected for culturing and miniprep and insertions were verified by Sanger sequencing. Photinuspyralis Luciferase was fused directly to the C terminus of Cry1 and Cry2 in the pCMV-Tag3C-Myc vector. Full-length Cry2 was inserted at the exact location in pMU2-P(Cry1)-(intron336)-Myc where Cry1 was deleted.

HEK-293A cells were seeded in 6-well tissue culture plates at a density of 4 × 10⁵ cells per well. Cells were transfected the same day with FuGENE 6. The following constructs were used for all transfections: p3X-FLAG-CMV-10-mBmal1, p3X-FLAG-CMV-10-mClock, pCMV-Myc-Cry1²¹, pCMV-Myc-Cry2²¹, and pcDNA3.1-mPer2-V5⁷⁴. For IPs involving all four clock components, cells were transfected with 45 ng of mBmal1, 1.5 μg of mClock, 600 ng of mPer2, and 150 ng of either Cry1, Cry2, or mutants thereof, in a total of 2.295 μg of DNA. Empty vector (pcDNA3.1-A) was used to even out DNA cocktails. For IPs of only BMAL1, CLOCK, and CRY, cells were transfected with 45 ng of mBmal1, 1.5 μg of mClock, and 150 ng of Cry plasmid in a total of 1.695 μg of DNA. Finally, for IPs of PER and CRY alone, cells were transfected with 600 ng of mPer2 or 300 ng of Cry in a total of 900 ng of DNA. After 48 h, cells were lysed in 200 μL of TGED buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 5% glycerol, 0.5 mM dithiothreitol, 0.5% Triton X-100, protease inhibitor (1:50)) for 30 min before a 10 min centrifugation to remove cellular detritus. Ten percent of the supernatant solution was saved for use as input and the remainder was incubated for 3 h with 40 μL of anti-MYC-conjugated beads to immobilize MYC-CRY. Beads were washed with 1 ml of TGED buffer twice. Protein was released from beads by boiling in 50 μL of sodium dodecyl sulfate (SDS) sample buffer (26.3 mM Tris-HCl, pH 6.8, 4.2% glycerol, 0.84% SDS, 10.5% β-mercaptoethanol, 0.21 mg/mL bromophenol blue) and analyzed by immunoblot using anti-MYC, anti-V5, and anti-FLAG-HRP (horse radish peroxidase) for CRY, PER2, and CLOCK/BMAL1, respectively. Catalog numbers for the antibodies can be found in Supplementary Table 3. For immunoblots of all four proteins: anti-MYC was used at a concentration of 1:4000 on input blots and between 1:25 000 and 1:80 000 on IPs; anti-FLAG-HRP was used at a concentration of 1:200 000 on input blots and between 1:3000 and 1:100 000 on IPs; anti-V5 was used at a concentration of 1:2 000 000 on input blots and 1:10 000 000 on IPs. For immunoblots of CLOCK, BMAL1, and CRY: anti-MYC was used at a concentration between 1:3000 and 1:5000 on input blots and between 1:20 000 and 1:50 000 on IPs; anti-FLAG-HRP was used at a concentration between 1:10 000 and 1:50 000 on input blots and between 1:1000 and 1:5000 on IPs. For immunoblots of CRY and PER2: anti-MYC was used at a concentration of 1:5000 on input blots and 1:50 000 on IPs; anti-V5 was used at a concentration between 1:5000 and 1:25 000 on input blots and between 1:50 000 and 1:200 000 on IPs. The anti-mouse immunoglobulin G, HRP-linked secondary was always used at a concentration of 1:10 000. Blots were imaged on radiography film with Clarity ECL Substrate. Appropriate replicate size was determined empirically from extensive use of these assays in previous work²⁰.

HEK-293A cells were plated in 35 mm tissue culture dishes at a density of 4 × 10⁵ cells per well. Cells were transfected (FuGENE 6) same day with 500 ng of a Cry-Luc fusion construct (pCMV-Myc-Cry1-Luc or pCMV-Myc-Cry1-Luc) and 200 ng of an EGFP construct (pEGFP) as a transfection control. At 48 h after transfection, the medium was exchanged for LumiCycle recording medium supplemented with cycloheximide (100 μg/ml). Plates were transferred to the LumiCycle and bioluminescence was monitored from each dish continuously for ~70 s every 10 min using a photomultiplier tube at 37 °C. Data were corrected for background noise from the photomultiplier tube as described above. The first 10 h of recording were normalized to the first data point and half-life was determined by nonlinear, one-phase exponential decay analysis.

CRY alignments for Fig. 4 and Supplementary Fig. 3 were performed with CLC Main Workbench 7 using default settings for the slow alignment mode. Sequences were accessed and downloaded from NCBI.

Full-length Venus (Ex515/Em528) and Cerulean (Ex433/Em475) have been described previously⁷⁵. PCR-amplified Cry1 and Cry2 cDNAs were inserted into the NheI and AgeI sites upstream of the coding region of CerN and VenN. Full-length Clock cDNA was amplified by PCR with an N-terminal SacII site and a C-terminal blunt polymerization end and ligated into SacII and SmaI sites in CerC. Residues 2–88 and 396–855 of Clock were deleted from the construct by site-directed mutagenesis. Truncated Venus (VenN (1–155 amino acids)) and Cerulean (CerN (1–155 amino acids), CerC (156–239 amino acids)) fragments were previously generated in the pEGFP-C1 (Clontech) backbone by site-directed mutagenesis¹⁵.

For BiFC competition experiments, 25 ng H2B-mRFP1⁷⁶ was mixed with 200 ng CerC-Clock (89–395), 10 ng p3X-Flag-Bmal1, 200 ng of Cry1-VenN or Cry1-CerN, and 200 ng of Cry2-CerN or Cry2-VenN in a total of 735 ng of DNA. A total of 8 × 10⁴ HEK-293A cells were plated in each well of a 24-well black Visiplate. On the same day, cells were transfected with FuGENE 6. Plates were washed with phosphate-buffered saline (PBS) once 48–60 h after transfection, fixed with 4% paraformaldehyde in PBS for 15 min, and then immersed in PBS. A control well for background fluorescence was transfected with non-complementary BiFC vectors (200 ng Cry1-VenN, 200 ng Cry2-CerN, 25 ng H2B-mRFP1). Controls for normalization of the competition experiments were transfected with single complementation pairs (25 ng H2B-RFP1, 200 ng CerC-Clock (89–395), 10 ng p3X-Flag-Bmal1, and 200 ng of Cry2-CerN or Cry2-VenN in a total of 735 ng of DNA).

Fluorescence images were acquired on a Deltavision Personal DV Imaging System equipped with an inverted 10× (for quantification) or 20× (for publication images) 0.45NA UPLFL objective and a Microtiter stage on an Olympus IX71 microscope. Four locations in each well were selected and autofocused with red fluorescent protein 1 (RFP1) fluorescence. Single layer images for quantification or 20-layer Z stacks with 1 mm steps for deconvolution and presentation were scanned in the channel sequence of RFP (Ex575/25; Em632/60), YFP (Ex513/17; Em559/34), and CFP (Ex438/24; Em465/30) filter sets with the Z step first. Image stacks were deconvoluted with the Softworx deconvolution module and maximal intensity Z projections of layers were built.

For quantification of BiFC competition results, original image files were imported to and organized by ImageJ (NIH), and exported image sequences were pushed through a custom pipeline run by Cellprofiler. Briefly, the nuclei were first recognized based on RFP fluorescence, and then inverse-masked with strong aggregates in the Venus and Cerulean channels. Mean Venus and Cerulean intensity values of each masked nucleus were measured. About 40–120 cells were identified and measured in each image, and the average values of cells in one image were obtained for further normalization and statistical analysis. Background intensity values from Venus and Cerulean channels were subtracted from competition experiment intensity values. Background subtracted competition values were then normalized to background subtracted intensity values from single complementation wells. Four images from each well were averaged and the experiments were replicated three times.

Statistical parameters are reported in the figure Legends and indicated in the figures where appropriate. Unpaired t tests with Welch’s correction were performed by GraphPad Prism software. Welch’s t test was chosen to compare the period means and half-life means because we do not know whether the various mutant populations have equal variances to the WT protein. Protein half-lives were determined by normalizing each luminescence value of a given sample to the initial luminescence reading, and then fitting a one-phase decay curve to the resulting data in GraphPad Prism.

Custom scripts used in processing the SCA data, the alignment used for the SCA, the annotated SCA notebook, and the SCA database file are all available upon request. The custom data pipeline for CellProfiler is available upon request.

Other data are available from the corresponding author upon reasonable request.

Supplementary Information(PDF 7767 kb) Peer Review File(PDF 454 kb)