The Arg-293 of Cryptochrome1 is responsible for the allosteric regulation of CLOCK-CRY1 binding in circadian rhythm

Mammalian circadian clocks are driven by transcription/ translation feedback loops composed of positive transcriptional activators (BMAL1 and CLOCK) and negative repressors (CRYPTOCHROMEs (CRYs) and PERIODs (PERs)). CRYs, in complex with PERs, bind to the BMAL1/CLOCK complex and repress E-box – driven transcription of clock-associated genes. There are two individual CRYs, with CRY1 exhibiting higher affinity to the BMAL1/CLOCK complex than CRY2. It is known that this differential binding is regulated by a dynamic serine-rich loop adjacent to the secondary pocket of both CRYs, but the underlying features controlling loop dynamics are not known. Here we report that allosteric regulation of the serine-rich loop is mediated by Arg-293 of CRY1, identified as a rare CRY1 SNP in the Ensembl and 1000 Genomes databases. The p.Arg293His CRY1 variant caused a shortened circadian period in a Cry1 2 / 2 Cry2 2 / 2 double knockout mouse embryonic fibroblast cell line. Moreover, the variant displayed reduced repressor activity on BMAL1/CLOCK driven transcription, which is explained by reduced affinity to BMAL1/ CLOCK in the absence of PER2 compared with CRY1. Molecular dynamics simulations revealed that the p.Arg293His

The circadian clock modulates numerous behavioral and physiological processes through rhythmic transcriptional regulation (1,2). In humans, physiological parameters such as alertness, memory, heart rate, blood pressure, and immune responses are clock-controlled (3)(4)(5)(6). Additionally, genetic and epidemiological studies have linked clock disruption to various adverse metabolic phenotypes (7). Circadian rhythms at the metabolic and transcriptional level are conserved across phyla, influencing physiology in cyanobacteria, fungi, plants, and virtually all animals (8).
The cellular circadian clock is composed of positive and negative transcriptional feedback loops. Transcriptional factors BMAL1 and CLOCK, having basic helix-loop-helix (bHLH) and PER-ARNT-SIM (PAS) domains, are central to circadian biology. They form heterodimer and bind to E-box elements (CACGTG), where they initiate the transcription of clock-controlled genes including the Period (Per) and Cryptochrome (Cry) (9)(10)(11). PER and CRY form a heterodimer which translocates into the nucleus upon phosphorylation by casein kinase ID, and inhibits BMAL1/CLOCK-mediated transcription. Ultimately, PER and CRY act as negative regulators of their own transcription (12)(13)(14)(15). Although the initial transcriptional activation step generates the positive arm of the circadian clock, the latter inhibitory step generates the negative arm of the feedback loop (9,16). F-box proteins, b-transducin repeat containing protein 1 (beta-TrCP1), and F-box and leucine rich repeat protein 3 (FBXL3) ubiquitinate PER and CRY, and thus initiate their proteasomal degradation, respectively (17,18). Degradation of CRY and PER proteins relieves the repression on BMAL1/CLOCK, restarting the cycle anew (19).
There are two Cry genes in mammals: Cry1 and Cry2. They code for proteins that belong to the photolyase/cryptochrome family and act as transcriptional repressors (12). Several studies suggest that CRYs might have different roles in the circadian clock. First, studies with Cry1 2/2 and Cry2 2/2 mice suggest that each Cry plays a different role in animal behavior. Cry1 2/2 animals exhibit shorter free-running circadian rhythm, whereas Cry2 2/2 animals display a longer free-running circadian rhythm (14,20). Cry1 2/2 Cry2 2/2 animals show an arrhythmic phenotype under constant dark conditions (14,15). Second, CRY1 and CRY2 are present at different circadian times on chromosomal DNA (21). Third, when compared with CRY2, CRY1 is a more potent repressor of BMAL1/CLOCK transactivation on E-boxes (22,23). Finally, CRY1 specifically interacts with cystathionine b-synthase by regulating one-carbon and trans-sulfuration pathways (24). All these studies show functional differences between the CRYs in the circadian clock mechanism.
CRYs contain a photolyase homology region (PHR) consisting of primary (FAD binding) and secondary pockets and extended C-terminal region (25,26). PHR and C-terminal regions of the CRYs are required to maintain rhythmicity and amplitude of circadian rhythm, respectively (22,27). CRY1 interacts directly with CLOCK and BMAL1 through two distinct regions on the PHR domain (28). C-terminal regions of CRYs interact with the transactivation domain (TAD) of BMAL1 (29,30). Biochemical and computational modeling analyses show that the HI loop on the PAS-B domain of CLOCK docks into the secondary pocket of CRYs (28). Moreover, mutational analysis of CRYs indicated that residues around the secondary pocket of CRY1 and CRY2 are critical for the CLOCK binding and their repression activity (31). In fact, a human gain-of-function CRY1 variant (exon 11 skipping mutation in C-tail of CRY1) found in people suffering from familial delayed sleep phase disorder and attention deficit/hyperactivity disorder exhibits high affinity to BMAL1/ CLOCK (32,33).
Coiled-coil helix in the C-terminal of CRYs, which interacts with PER2 and plays a role in CRY nuclear shuttling, inhibits the binding of FBXL3 and subsequent proteasomal degradation (34,35). FBXL3 competes with PER2 to interact with CRYs by wrapping around the coiled-coil helix via b-sheet domain. In addition, the last five C-terminal residues of FBXL3 penetrate to the FAD binding pocket which, in turn, facilitates the ubiquitination of CRYs (36,37). Crystal structure of CRY2 PHR-PER2 C-terminal CRY-binding domain (CBD) revealed an unusual binding mode of PER2-CBD onto CRY2 (34). The N-terminal half of PER2-CBD reaches the rim of secondary pocket of the CRY2-PHR. This interaction allows the serine loop, located adjacent to the secondary pocket, to adapt different conformations. A recent study indicates that PER2 remodels the serine loop of CRY2 and, in turn, enhances its affinity toward BMAL1/CLOCK (38). In the same study, replacement of amino acid residues in the serine loop of CRY1 reduced its affinity to BMAL1/CLOCK. However, communication and allosteric regulation between primary and secondary pockets of CRY1 have not been reported so far.
Here, we characterized three SNPs (p.Gly144Val, p.Arg293His, and p.Arg348Cys) identified from 1000 Genomes project and Ensembl databases. Among them, analysis of the p.Arg293His CRY1 variant indicated that it had a longer t 1/2 and weaker repression activity and caused shorter period of circadian rhythm. Further biochemical analysis showed that this variant had a weaker affinity toward the BMAL1/CLOCK complex in the absence of PER2. Interestingly, a comparison of CRY2 and the p.Arg293His CRY1 variant showed very similar molecular characteristics in terms of t 1/2 and binding properties to BMAL1/ CLOCK. However, in rescue assays using Cry1 2/2 Cry2 2/2 mouse embryonic fibroblasts CRY2 and p.Arg293His CRY1 showed some differences. The period length of rhythm observed in p.Arg293His CRY1 rescue cell lines were shorter than WT CRY1, but longer than CRY2. To understand all these changes mediated by Arg-293 within the structural context of CRY1, we performed MD simulations. The MD studies of both mutant and WT CRY1 suggest that Arg-293 is involved in allosteric regulation of CLOCK binding by regulating the dynamicity of the serine loop adjacent to the secondary pocket.
Collectively, this study provides new insights into the circadian clock mechanism. In particular, our study suggests potential new targets for drug discovery by identifying novel amino acids affecting the repressor and binding activities of the CRY1 and the possible allosteric regulation in the CRY1 secondary pocket.

Selection of CRY1 SNPs
Several SNPs with either low or high frequencies are associated with different types of diseases. However, the functional significance of rare missense mutations deduced from largescale genomic studies and how these mutations affect protein function are largely unknown. Because of that, we investigated the effect of missense variants of the human CRY1 gene. The SNPs for CRY1 were selected from the 1000 Genomes project and Ensembl databases (39, 40) based on in silico analyses and their location within the CRY1 structure. Among 12,000 CRY1 SNPs, a majority of them are located within noncoding regions. Of the SNPs reported within the CRY1 coding region (116 missense mutations are detected), we have focused on missense mutations affecting the PHR domain of the CRY1, which is crucial for its function (22). Three of these variants predicted to be detrimental to protein structure or function were selected based on their SIFT, PolyPhen, and Provean scores (Table 1). These variants were interpreted as "disease-causing" by Muta-tionTaster (41). Additionally, VarSome-Verdict and ClinVar were used to understand the clinical effects of these variants. Although these variants were classified as "variant of uncertain significance-VUS", and all of them meet two pathogenic American College of Medical Genetics and Genomics criteria (guidelines for interpretation of sequence variants) (42). None of these variants were reported in ClinVar. The p.Arg293His and p.Arg348Cys CRY1 variants were detected as heterozygous; the genotype of the p.Gly144Val was not reported in Ensembl. To analyze the degree of conservation of these SNPs, we performed multiple sequence alignment (Fig. 1A). The Arg-293 in CRY1 is conserved in all CRY1, whereas Gly-144 is conserved in all CRY1 proteins with the exception of mosquito CRY1. Finally, Arg-348 is conserved in CRY1 from most organisms, but not in CRY1 from sponge, mosquito, housefly, green sea turtle, blowfly, and stony coral. WT amino acids in all three SNPs are highly conserved in the CRY1 across different organisms, which suggests a potential role in protein function (Fig.  1A). Thus, we aimed to investigate the effects of these variants at the molecular level. To do that, we generated mutant constructs of mouse Cry1 (mCry1) by site-directed mutagenesis, and these variants were subjected to further functional analysis.

The effect of SNPs on CRY1 t 1/2 and molecular clock function
The functional characteristics of the selected rare missense variations were studied in detail using biochemical and cellbased assays. Mapping of SNPs onto the crystal structure of CRY1 indicated that these SNPs are located on functionally important domains. Gly-144 and Arg-348 are both exposed to solvent and are located on the loop between helices a5 and a6, or in helix a14, respectively (Fig. 1B). Arg-293 is located within the FAD binding pocket and FBXL3 interaction site which is adjacent to the secondary pocket (Fig. 1, B and C). Because FBXL3 is the E3 ligase responsible for ubiquitination of CRY1, mutations in this region of FBXL3 might change the stability of CRY1 (Fig. 1D). We, therefore, investigated the effect of these mutations on the t 1/2 of CRY1 by monitoring the degradation of CRY1 fused to LUC (CRY1::LUC) at the C-terminal. After generating Cry1 variants, plasmids were transiently transfected into HEK293T cells, which were then treated with cycloheximide to inhibit protein synthesis. Luminescence was monitored to determine the t 1/2 of the proteins. Stability of the p.Arg348Cys CRY1 and p.Gly144Val CRY1 variants were found to be similar to WT CRY1 ( Fig. 2A). However, the t 1/2 of the p.Arg293His CRY1 variant was significantly longer than that of WT CRY1 and shorter than luciferase. We further confirm this result with biochemical assays. After transfection of HEK293T cells with proper plasmids, cells were treated with cycloheximide and harvested at the 4th and 8th h. Analysis of samples with Western blotting indicated that p.Arg293His CRY1 is indeed more stable than the WT CRY1 ( Fig. 2A, right panel). These results suggested that although mutations distant from FBXL3 interaction sites did not change the stability of CRY1, mutation in the binding region of FBXL3 (i.e. p.Arg293His) increased the t 1/2 of CRY1 probably because of the attenuated interaction between CRY1 and FBXL3.
We next determined the effect of these variants on circadian clock function. To do this, we employed a complementation assay using Cry1 2/2 Cry2 2/2 mouse embryonic fibroblast cells (DKO-MEF). Plasmid constructs carrying WT Cry1 were shown to rescue circadian rhythms in DKO-MEF when expressed under the control of mCry1 promoter containing E/E9-box and D-box elements and the ROR/REV-ERB-binding element in the first intron (43). Thus, we introduced the p.Gly144Val, p.Arg293His, and p.Arg348Cys variants into the pMU2-Cry plasmid. The Cry expression vector carrying either the variants or WT Cry1 and the mPer2-dLuc reporter plasmid were transiently expressed in DKO-MEF cells. We monitored bioluminescence for 5 days. Although WT and variants were able to restore the circadian rhythm, they had varying effects on period length. The period length for WT CRY1 was calculated as 26.3 h (Fig. 2B). The p.Gly144Val and p.Arg348Cys CRY1 variants increased the period length to 27.8 and 27 h, respectively. Despite the longer t 1/2 of the p.Arg293His CRY1 variant, the mutant caused a significantly shorter period length (23.8 h) and displayed weaker repressor activity based on higher luminescence intensities compared with WT CRY1. Because rhythm is generated as a result of interaction between core clock proteins, e.g. CRY1, PER2, BMAL1, and CLOCK, this intriguing observation on p.Arg293His mutant led us to analyze binding properties of these mutants to the core clock proteins.
p.Arg293His CRY1 binding to BMAL1/CLOCK is differentially disrupted To assess binding of each CRY1 variants to CLOCK and BMAL1, HEK293T cells were transiently transfected with expression plasmids coding for CLOCK, Bmal1, and either WT or Cry1 variants. The CRY1-myc-His (variants or WT)containing protein complexes were then precipitated using Ni-NTA agarose, and the samples were probed for the presence of CLOCK and BMAL1 by Western blot analysis as described in Ref. 31. To assess the binding of CLOCK properly, comparable amount of the CRYs were used in pulldown assay. The p.Gly144Val and p.Arg348Cys CRY1 variants exhibited similar affinities to BMAL1/CLOCK compared with WT CRY1, whereas the p.Arg293His CRY1 variant exhibited significantly less affinity to BMAL1/CLOCK (Fig.  3A). A similar experiment was performed in the presence and absence of PER2 to investigate the possibility that PER2 is required for proper binding. All three CRY1 variants and WT CRY1 bound similar amounts of PER2. Nevertheless, the p.Arg293His CRY1 variant was to bind less BMAL1/ CLOCK in the absence of PER2 (Fig. 3B). Biochemical binding analysis showed that reduced repression by the p.Arg293His variant of CRY1 is most likely caused by attenuated binding to the BMAL1/CLOCK dimer. However, the p.Gly144Val and p.Arg348Cys variants retained binding activity to the BMAL1/CLOCK dimer. Overall, the p.Arg293His CRY1 variant had longer t 1/2 , rescued circadian rhythms with shorter period in a weakly repressed state in DKO-MEF cells (Fig.  3B) and exhibited reduced affinity for the BMAL1/CLOCK dimer even in the presence of PER2 (Fig. 3). These features of the p.Arg293His CRY1 variant reported here are very similar to those reported for CRY2 (22,31), which led us to perform additional studies.

p.Arg293His CRY1 variant exhibits similar characteristics to CRY2
The secondary pocket of CRY1 is critical for its interaction with CLOCK. Subtle changes in this pocket can result in substantial differences in periodicity in cycling cells (28,38). For example, differential rhythms driven by CRY1 and CRY2 are the result of these changes (23). To directly compare p.Arg293His CRY1 and WT CRY2, we measured their halflives and effects on circadian rhythms using DKO-MEF cells. Both the pArg293His CRY1 variant and WT CRY2 exhibited similar half-lives (Fig. 4A). It has been shown that CRY2 can rescue circadian rhythms when the pMU2 plasmid with Cry2 cDNA is transfected in DKO-MEFs (31,43). Therefore, we cloned Cry2 cDNA into the pMU2 plasmid and transfected this construct into DKO-MEF to see whether it rescues circadian rhythms. Compared with CRY1, CRY2 was able to rescue circadian rhythms with a shorter period, which was consistent with a previous study (31). Our results showed that cells expressing CRY2 had higher overall luminescence intensity than cells expressing CRY1, suggesting higher BMAL1/CLOCK activity, again consistent with a previously published report (22). Interestingly, cells expressing the p.Arg293His CRY1 or CRY2 had similar luminescence level (Fig. 4B). In the complementation assay, DKO-MEF cells expressing WT CRY1 or CRY2 had h, which is shorter than that of WT CRY1 and longer than CRY2 (Fig. 4B). Finally, the amplitude of the circadian rhythm was higher in cells expressing p.Arg293His CRY1 variant and CRY2 than WT CRY1 (Fig. 4B). To compare the ability of p.
Arg293His CRY1 and CRY2 to repress BMAL1/CLOCK transactivation, we used an mPer1-dLuc assay using Cry1 2/2 Cry2 2/2 MEF cells. These results indicated that both CRY2 and p.Arg293His CRY1 had comparable levels of repressor activity that was dose-dependent and statistically significantly different from that of WT CRY1 (Fig. 4C). Additionally, binding of  WT and p.Arg293His CRY1 and CRY2 to BMAL1, CLOCK, and PER2 were analyzed side by side (Fig. 4D). We performed pulldown assay in the absence and presence of PER2 to assess the binding ability of WT CRY1, CRY2, and p.Arg293His CRY1 to CLOCK. All CRYs tested in this study had comparable binding to CLOCK in the presence of PER2 (Fig. 4D). However, CRY2 and p.Arg293His CRY1 had attenuated binding to CLOCK compared with WT CRY1 in the absence of the PER2. Collectively, our results suggested that the p.Arg293His CRY1 variant possesses similar properties to CRY2 in terms of stability, binding to core clock components, repression activity and rescuing circadian rhythms in DKO-MEF, except p.Arg293His CRY1 generated longer rhythm than CRY2.

Understanding the role of Arg-293 in CRY1 function
Although p.Arg293His CRY1 variant exhibits CRY2-like in vitro properties, the conservation of this amino acid residue in all mammalian CRY1s suggests that it might play an important role, such as regulation of CLOCK binding to the secondary pocket of CRY1. To investigate this, we employed MD simulations using CRY1 structural data. Recently, it has been shown that CRYs bind to the CLOCK PASB domain through the secondary pocket, and this is regulated by the serine loop (38). We, therefore, explored possible dynamical differences between WT CRY1 and p.Arg293His CRY1 around the secondary pocket via MD simulations. p.Gly144Val CRY1 and p.Arg348Cys CRY1 structures were also simulated as controls. For this purpose, we utilized mouse CRY1 crystal structure to generate a complete PHR domain of the individual variants. Each protein was sampled for 300 ns, and the root mean square deviation (RMSD) values were determined to show that simulations reached equilibrium (Fig. 5A). To investigate the dynamics of each residue, root mean square fluctuation (RMSF) values were assessed (Fig. 5B). We observed differences in the RMSF values of residues in the serine loop (residues [38][39][40][41][42][43][44][45][46][47][48]. When compared with WT CRY1, the residues in p.Arg293His CRY1 simulations showed minimal dynamism/fluctuation (Fig. 5B). Because the serine loop regulates the volume and access to the secondary pocket, we analyzed the volume of the secondary pocket using POVME 3.0 software. Analysis showed that simulations of p.Arg293His CRY1 sampled significantly smaller secondary pocket volume than WT CRY1 (Fig. 5C). Secondary pocket volume of p.Gly144Val CRY1 was comparable to WT CRY1. Interestingly, p.Arg348Cys CRY1 simulation sampled a larger secondary volume pocket. Because secondary pocket volume of WT CRY1 allows CLOCK binding, we do not expect larger pocket volume will affect CLOCK binding to p.Arg348Cys CRY1. In fact, rescue assays (Fig. 2B) and BMAL1/CLOCK binding analysis to CRY1 variants (Fig. 3) confirmed that p.Arg348Cys mutation did not change the binding of CRY1 to CLOCK. To further examine the relative internal dynamics of the loop, we analyzed the distance between the C a atoms of Gly-43 located on the loop and Phe-105 (Fig. 5, D and E). The average loop distances of CRY1 and p.Arg293His CRY1 were not significantly different. However, the variance of loop distances sampled throughout the simulations were significantly different (Levene's test). To compare the scatter of loop distance and its correlation with the volume of the secondary pocket, we plotted loop distance against the secondary pocket volume (Fig. 5E). Our analysis of secondary pocket volume and loop distance showed that p.Arg293His CRY1 behaved similarly to WT CRY1 (Fig. 5, C-E). Despite this similarity between CRY1 and p.Arg293His CRY1 simulations, p.Arg293His CRY1 only sampled a fraction of the space explored by WT CRY1 (Fig. 5E). In line with secondary pocket volume analysis, p.Gly144Val sampled similar loop distance to WT CRY, p.Arg348Cys sampled longer loop distance. These observations in MD simulations implied that the dynamics of the serine loop in CRY1 and p.Arg293His CRY1 are regulated differently. Although the serine loop in WT CRY1 is more dynamic, reduced dynamicity of this loop in p.Arg293His CRY1 might explain the attenuated binding of p.Arg293His CRY1 to CLOCK. To explain this differential regulation on the serine loop in detail, we analyzed possible communication pathways between Arg-293 and the serine loop by using Weighted Implementation of the Suboptimal Paths (WISP) software as described in Ref. 44. The WISP algorithm can calculate both optimal and suboptimal (longer) communication pathways between nodes (residues). Residue 293 was used as a source residue, and several possible residues on the serine loop (Gly-43, Ser-44, Ser-45, Asn-46) were used as a sink to determine promising pairs for communicating pathways. A total of 500 pathways between residues Arg-293 and Gly-43 for each WT and p.Arg293His CRY1 were calculated. The shortest pathway length for p.Arg293His CRY1 was calculated as 3.26 Å. The shortest communicating pathway involves the His-293, Ile-392, Ser-391, Trp-390, Gly-43 residues. However, the length of shortest pathway for WT CRY1 was calculated as 5.19 Å, which involves the Arg-293, Phe-296, Ala-300, Pro-104, Gly-106, Arg-109, Trp-8, Ile-36, Asp-38, Gly-43 residues (Fig. 5, F and G). Not only is the shortest pathway changed, but also the suboptimal pathways sampled longer lengths in p.Arg293His CRY1 (Fig. 5G). Residues involved more than 200 suboptimal paths between Arg-293/ His-293 and Gly-43 are Asp-38, Pro-104, Gly-106, Ile-36, Ala-300, Arg-109, Phe-296, Trp-8, and Ala-299 for WT CRY1; Glu-382, Phe-296, Gln-393, and Ile-392 for p.Arg293His CRY1. Both representation of pathways and histogram of . Statistical analysis was performed using an unpaired t test with a Welch's correction (***, p , 0.0001; **, p , 0.005; *, p , 0.05; WT CRY1 versus R293H, p = 0.0226; WT CRY1 versus CRY2, p = 0.0034; R293H CRY1 versus CRY2, p = 0.4854). Degradation of WT CRY1 and R293H CRY1 was analyzed by immunoblot using anti-Myc for CRYs and anti-b-actin for loading controls (right panel). Degradation of LUC was compared with WT and R293H CRY1::LUC. B, pArg293His CRY1 and CRY2 rescued the rhythm similarly, although periods are still significantly different. Rescue assays were performed as described in Fig. 3. Representative rhythms of three independent experiments are shown. Periods and amplitudes of all plates were reported (4-WT, 6-R293H CRY1, and 6-CRY2). Significance analysis was performed using unpaired t test with Welch's correction (***, p , 0.0001; **, p , 0.005; *, p , 0.05; WT CRY1 versus R293H, p , 0.0001; WT CRY1 versus CRY2, p , 0.0001; R293H CRY1 versus CRY2, p , 0.0001). C, repression analysis. Cry1 2/2 Cry2 2/2 MEF cells were transfected with plasmids encoding Bmal1 (50 ng), Clock (125 ng), and different amounts of WT Cry1, Cry2, and Cry1R293H, along with the pGL3-mPer1:luc (50 ng) reporter plasmid and pRL-TK Renilla (4 ng) as a transfection control. Comparisons were made between WT CRY1 and CRY1R293H and WT CRY1 and CRY2 (***, p , 0.0001; **, p , 0.005; *, p , 0.05). D, pulldown analysis of WT or pArg293His CRY1, CRY2 with BMAL1, CLOCK, and PER2. HEK293T cells were transfected with plasmids expressing Bmal1-FLAG, CLOCK-FLAG, and Cry-His-Myc. Proteins were pull downed with Ni-NTA agarose resin and analyzed with Western blotting. This blot is the representation of three independent experiments.

Allosteric regulation in Cryptochrome1
pathway lengths showed that the mutation introduced new communicating pathways between His-293 and Gly-43. Taken together, these computational findings suggest that the serine loop is allosterically regulated by residue Arg-293 in CRY1 and reduces the dynamicity of the serine loop, keeping it in a closed conformation. Our search in the ClinVar database revealed that there was no disease phenotype associated with this variant.

Discussion
Mammalian CRYs are integral components of the circadian clock mechanism and are strong transcriptional repressors of BMAL1/CLOCK transactivation (45)(46)(47). Several genetic studies have highlighted the relationship between CRY variants and different types of diseases. For example, CRY1 variants have been associated with depression and mood disorders (48)(49)(50), elevated blood pressure and hypertension (49), and insulin resistance (51,52). This clinical heterogeneity implies a variety of different functions for CRY1 potentially caused by different functional domains of the protein. Multiple sequence alignment shows that CRYs have variable extended C-terminal domains that range from 30 to 300 amino acids depending on the species (25,26). N-terminal domain exhibits high homology to photolyases and is called the PHR domain.
The C terminus of mammalian CRY possesses a highly conserved helix that has a periodic spacing of nonpolar residues common to coiled-coil and is called the coiled-coil helix (53). In mammals, this region is competitively bound by FBXL3 (54) and PER2 (35) to regulate the stability of CRYs, and to tune the circadian clock machinery. The PHR domain essentially consists of two important regions, an a-helical domain (primary pocket), and an a/b domain (secondary pocket). The a/b domain (secondary pocket) has been shown to be important for the interaction with the CLOCK PAS B domain and maintains the BMAL1/CLOCK transcription factor in a repressed state to close the circadian feedback loop (28).
Although rare CRY1 variants have been previously shown to be associated with several disorders, the underlying mechanisms are yet to be characterized. Here, we not only investigated the potential functional impact of rare nonsynonymous SNPs on the function of CRY1 but also revealed the allosteric regulation between primary and secondary pockets of CRY1 by characterizing one of the rare mutants. The three variants, which map to important domains of the CRY1 and are predicted to affect its function, were selected from Ensembl and the 1000 Genomes project. Arg-348 and Gly-144 residues map to the surface; Arg-293 is located within the primary pocket, FAD-binding domain, and FBXL3 binding site of CRY1 (Fig. 1). We initially analyzed the effect of mutations on the t 1/2 CRY1 protein using the CRY1::dLUC degradation assay. p.Arg348Cys and p.Gly144Val CRY1 variants had no effect on the stability of CRY1, whereas the p.Arg293His variant significantly increased the stability compared with WT CRY1 (Fig. 2A). Next, we utilized a cell-based rescue assay in Cry1 2/2 Cry2 2/2 MEF cells with a special Cry1 rescue plasmid to identify their effects on the circadian rhythm. Surprisingly, p.Arg293His CRY1 variant had longer t 1/2 than the WT CRY1 and caused shorter period length with enhanced luminescence. This implies less repression activity (Fig. 2B), which is the indicator of weak binding of mutant CRY1 to BMAL1/CLOCK. Other two variants caused slight increase in the period length, which is consistent with previous findings indicating that mutation of residues on the surface of the a-helical domain have a tendency to change the repressor activity of mammalian CRYs (55). Despite earlier studies suggested that as the stability of CRY increases by either mutation or chemicals, period of circadian rhythm becomes longer; however, recent findings unveiled that not the stability but the ability of CRY to bind BMAL1/CLOCK is the primary factor in determining the period length (30,31,33,56).
To test binding of CRY to core clock proteins, pulldown experiment was performed with CRY1 and BMAL1/CLOCK in the presence and absence of PER2. As rescue assay results suggested, p.Arg293His CRY1 has deficiency to bind CLOCK, even in the presence of PER2 (Fig. 3). On the other hand, p.Gly144Val and p.Arg348C mutants showed comparable binding to core clock proteins with WT CRY1. Interestingly, functional characterization of p.Arg293His, which is located in the a12-helix and possibly interacts with the secondary pocket, revealed different properties than the other CRY1 variants. That is to say, t 1/2 of the p.Arg293His CRY1 was longer than WT CRY1. Second, functional complementation with this variant in DKO-MEF resulted in a short circadian period phenotype with increased luminescence. Third, biochemical studies showed that this variant had reduced affinity to CLOCK. Interestingly, these features are more similar to CRY2 than CRY1 (31). Thus, we tested WT and p.Arg293His CRY1 and WT CRY2 in the same experimental set-up to determine the differences and similarities among these proteins. Although p.Arg293His CRY1 and WT CRY2 had comparable t 1/2 (Fig.  4A), repression capacity, and binding to BMAL1/CLOCK in pulldown assays (Fig. 4, C and D), p.Arg293His caused a longer rhythm than WT CRY1 but shorter than WT CRY2 (Fig. 4B). Despite similarities between p.Arg293His CRY1 variant and WT CRY2 in various aspects, rescue assay exhibited that they have slight functionality differences in the circadian clock mechanism (Fig. 4B).
To understand the dramatic change in the function of CRY1 upon Arg-293 to His-293 mutation and to assess the role of Arg-293 in CRY1, we employed molecular dynamics simulations. Computational analysis suggested that Arg-293 might be a critical residue for the regulation of the serine loop in CRY1, which is known to be important for CLOCK binding to the secondary pocket (38). Replacing Arg-293 with His changes a highly dynamic serine loop into a less dynamic state and closes the gate of the secondary pocket (Fig. 5, B-E), which may be re-sponsible for the reduced binding of CLOCK to the mutant CRY1. Further analysis of our MD simulations of both variant and WT CRY1 in terms of the communication pathways suggested that mutation of Arg-293 to His changes the path to serine loop, which shows the allosteric regulation of secondary pocket availability by the primary pocket (Fig. 5, F and G).
The rare variants characterized in this study were selected from published human SNPs, which are all in heterozygous state and, to our knowledge, have not yet been associated with any disease phenotype. It is possible that these variants may exhibit a disease phenotype in recessive state. Their effects can be further analyzed in mouse models by reverse phenotyping studies. In conclusion, we characterized the effect of rare human CRY1 SNPs on circadian clock function at the molecular level. Our findings suggest that Arg-293 is important for the allosteric regulation in CRY1 and has impact on the molecular clock.

Site-directed mutagenesis
For site-directed mutagenesis, Phusion polymerase-based quick-change method was used. Primers were designed using the recommended guidelines (57). The PCR reaction mixtures contained 0.3 mM dNTP, 5 ml of 103 Phusion GC Buffer (Thermo Scientific), 3% DMSO, 1 mM each primer, 30 ng of the template plasmid (mCry1 in pcDNA4A), and 1 unit of Phusion DNA polymerase in a 50 ml of final volume. The reaction conditions were set to 98°C for 15 s, 66°C for 30 s, and 72°C for 4 min, 20 cycles. PCR products were visualized in 1% agarose gel. The samples with correct-sized bands were digested with 1 unit of FastDigest DpnI enzyme (Thermo Scientific) for 1 h at 37°C and then 5 ml of sample was transformed to Escherichia coli DH5a cells. Colonies were picked to culture and then plasmids were isolated via MiniPrep (Macherey Nagel). Sanger sequencing (Macrogen) was used to confirm the mutations.

Cry-Luc construct generation
To form Cry-Luc constructs, coding sequence of firefly luciferase in pG5luc plasmid (Addgene) was amplified with primers having NotI and XhoI flanking sites. Product size was verified by visualizing in agarose gel and product was isolated from the gel by using NucleoSpin PCR and Gel Purification Kit (Macherey Nagel). Purified fragments were cloned to pJET1.2/ blunt vector (Thermo Scientific) using CloneJET PCR Cloning Kit (Thermo Scientific). 5 ml of sample was transformed to E. coli DH5a cells. Colonies were picked to culture and then plasmids were isolated via MiniPrep. Luc plasmid was double digested with NotI and XhoI; Cry pcDNA plasmids (WT or variant) were double digested with EcoRV and NotI FastDigest enzymes (Thermo Scientific). Cry plasmids were treated with FastAP (Thermo Scientific) to prevent self-annealing. After gel isolation the inserts and the destination vectors were cleaned and ligated by using T4 DNA ligase (Thermo Scientific). 5 ml of sample was transformed to E. coli DH5a cells. Colonies were picked to culture and then plasmids were isolated via MiniPrep. Presence of Cry and Luc were verified by double digestions with EcorV/NotI and NotI/XhoI, respectively.

Repression assay of BMAL1/CLOCK-dependent transactivation
Plasmids used in this study were gifted to us from Professor John Hogenesch from Cincinnati Children's Hospital. A mixture of 50 ng pSport6-Bmal1, 125 ng pSport6-CLOCK, 50 ng pGL3-mPer1::dLuc, 1 ng pRL-TK, 2.5 ng or lower doses of pcDNA4A-Cry1 (WT or variant), and empty pSport6 (to equalize the transfected amount of DNA) was reverse transfected to 4 3 10 4 HEK293T cells/well in a 96-well opaque plate via PEI transfection reagent. To determine the activity of BMAL1/ CLOCK-dependent transactivation without a repressor, the mixture was supplemented with empty pcDNA4/myc-His A instead of Cry1 construct. In each independent experiment, transfections were triplicated for each condition. The plates were incubated for 24 h at 37°C, 5% CO 2 . Firefly luciferase and renilla luciferase expression was determined using the Dual-Glo Luciferase Assay System (Promega) using manufacturer's protocol.
Pulldown assay 4 3 10 5 HEK293T cells (per well) were seed to 6-well tissue plate 24 h before the transfection. Cells were transfected with Cry1-His-Myc (WT or variants) and Cry2-His-Myc in pcDNA4-A with FLAG-CMV-Bmal1, FLAG-CMV-CLOCK via PEI transfection reagent for IP with BMAL1, CLOCK, and CRY1 as described in Ref. 57. FLAG-PER2-CMV plasmid was also transfected along with BMAL1, CLOCK, and Cry1 plasmids to pull down four clock proteins. For negative control BMAL1, CLOCK or BMAL1, CLOCK, and PER2 were transfected with empty pSport6 plasmid instead of Cry1. 24 h after transfection, cells were harvested via icecold PBS. After centrifugation pellets were lysed in 300 ml passive lysis buffer (PLB) (15 mM HEPES, 300 mM NaCl, 5 mM NaF, 1% Nonidet P-40 supplemented with fresh protease inhibitor) for 20 min on ice. To get rid of cell debris, samples were centrifuged for 15 min at 13,000 3 g at 14°C. 10% of the supernatant was saved as input. 15 ml Ni-NTA Agarose resin (Qiagen) per sample was equilibrated by washing two times with TBS-300 (15 mM Tris, 300 mM NaCl) supplemented with 25 mM imidazole and two times with PLB. Remaining supernatant was added on to equilibrated resins with 25 mM imidazole. Cell lysates and resins were mixed for ;2 h at 4°C to pull down CRYs. Resins were washed four times with TBS-300 (300 ml) having 35 mM imidazole. Proteins were isolated from resins by boiling in Laemmli Buffer (31.5 mM Tris-HCl, pH 6.8, buffer 10% glycerol, 1% SDS, 0.005% Bromphenol Blue, freshly added b-mercaptoethanol (5%)).

CRY-LUC degradation assay
40 ng of expression vector (Cry-Luc plasmid) was reverse transfected to 4 3 10 4 HEK293T cells on opaque 96-well plate with flat bottom via PEI transfection reagent. 48 h of posttransfection cells were treated with luciferin (0.4 mM final) and HEPES (10 mM final and pH 7.2). After 2 h, cycloheximide (20 mg/ml final) was added to wells to stop the protein synthesis. Plate was sealed with optically clear film and placed to Synergy H1. Luminescence readings were recorded every 10 min at 32°C for 24 h. t 1/2 of protein was calculated via one-phase exponential decay fitting function in GraphPad Prism5 software.
CRY degradation assay 500 ng of WT CRY1 or p.Arg293His CRY1 plasmids in pcDNA4A-His-Myc vector were transfected to HEK293T cells on a 6-well plate with PEI transfection reagent. 24 h after transfection cells were treated with cycloheximide (100 mg/ml final) to stop protein synthesis. Cells were harvested at the time of cycloheximide treatment (0 h) after 4 and 8 h. Levels of proteins were analyzed via Western blotting by using anti-Myc or antib-actin to detect CRY and b-actin levels, respectively. Details of antibody treatments are as explained in the pulldown assay section above.
Real-time bioluminescence rescue assay 3 3 10 5 Cry1 2/2 /Cry2 2/2 mouse embryonic fibroblasts (DKO-MEFs) were seeded onto 35-mm clear tissue culture plates (cells were obtained from Prof. Ueda's group). Cells were transfected with 4000 ng pGL3-Per2-dLuc (luciferase reporter) and 150 ng of the Cry1 expression vector (pMU2-P(CRY1)-(intron 336), obtained from the Ueda group) (WT or variant), using FuGENE6 transfection reagent according to the manufacturer's protocols. After 72 h, cells were synchronized by treatment with 0.1 mM dexamethasone for 2 h. Then, the medium was replaced with bioluminescence recording medium (1% DMEM powder (w/v), 0.035% sodium bicarbonate, 0.35% D(1) glucose powder, 0.01 M HEPES buffer, 0.25% penicillin/ streptomycin, 5% FBS), in which luciferin was freshly added (0.1 mM final concentration). Plates were sealed with vacuum grease and placed into a LumiCycle (Actimetrics). Bioluminescence was monitored for 70 s every 10 min for 5 days. Luminescence values were recorded and processed using the LumiCycle Analysis software. The first 20 h, data were discarded from the analysis because of high transient luminescence upon medium change. Period and amplitude were calculated for each sample. Each independent experiment was carried out with duplicate samples.

Molecular dynamics simulations
The protein structures of CRY1 and CRY2 were obtained via homology modeling. Complete PHR domain sequences of CRY1 and CRY2 were obtained from UniProt (mouse CRY1 (P97784), mouse CRY2(Q9R194)). Sequences were submitted to the RaptorX web server to obtain protein structures for CRY1 and CRY2 (58). The structure of the CRY1 mutant R293H was also obtained by the same procedure. All structures were protonated with PROPKA using the PDB2PQR webserver (59). The CHARMM36m force field was used for simulations (60). All three structures were solvated in a rectangular box using TIP3P water molecules and neutralized with the appropriate amount of sodium and chloride ions in visual molecular dynamics (VMD) (61). Energy minimization was performed for 100,000 steps. Systems were gradually heated and equilibrated in NPT ensemble for 140 ps. Throughout equilibration, constraints on the proteins were gradually removed. After equilibration, a production run of 300 ns for WT, G144V, R293H, and R348C CRY1 was performed. Total of 1200 ns consisting of two independent production runs were performed for each protein. All production runs were performed at 310 K and under 1 atm pressure using a Langevin thermostat and a Langevin barostat, respectively. All simulations were performed using NAMD software on in-house, local machines (62). To prepare trajectories for analysis, all trajectories were aligned to the initial structure, and the first 10 ns of all simulations were discarded. RMSD and RMSF calculations were performed using VMD. Each of the refined trajectories were saved as two multiframe PDB files with 1 ns and 10-ps intervals for POVME and WISP, respectively, for further analysis. Calculation of the secondary pocket was performed using POVME 3.0 software (63). For all trajectories, the volume of the secondary pocket was calculated with a 1-ns interval. The Wilcoxon rank sum test was used to determine significance of the difference between the pocket volumes implemented in R (64). Allosteric pathway analysis was performed using command line WISP 1.1 software (65). Visualizations of pathways were performed using VMD software.

Data Availability
All data are contained within the manuscript.
Acknowledgments-We thank Emma Harris from the University of Edinburgh for critical reading of the manuscript. We are grateful to Prof. Hiroki Ueda and Andrew Liu for the generous gift of the Cry1 2/2 /Cry2 2/2 mouse embryonic fibroblasts and the Cry1 rescue vector. Finally, we thank Prof. Hogenesch for the generous gifts of clock-related plasmids and the support. Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.