Conserved structural motifs in PAS, LOV, and CRY proteins regulate circadian rhythms and are therapeutic targets

Abstract

Despite evolving independently in diverse organisms, circadian clocks ubiquitously employ period‐ARNT‐single minded (PAS) and cryptochrome (CRY) proteins as key regulators coupling environmental variables into circadian regulation. In these systems, we often observe complex gene duplication events and evolution of specialized function despite retaining high‐sequence identity. These specialized functions often have evolved from ancestral photoactive proteins (LOV/CRY) where upon the ancestral photoactive ligand‐binding pockets have been co‐opted as protein–protein interaction motifs and targets for drug discovery. In this review, we dissect structural, biochemical, and computational studies of the PAS and CRY superfamilies within circadian clocks to highlight their molecular mechanisms and factors that position them as drug targets for diverse disease phenotypes. Particular focus is placed on discussing how photoactive members of the protein families can inform on allosteric mechanisms that couple cofactor‐binding sites to regulation of flexible signaling motifs relevant to circadian regulation and drug discovery.

Cryptochrome and PAS/LOV proteins play intricate roles in circadian clocks where they act as both sensors and mediators of protein–protein interactions. Their ubiquitous presence in signaling networks has positioned them as targets for small‐molecule therapeutics. This review provides a structural introduction to these protein families. Created in BioRender. Lester, A. (2025) https://BioRender.com/jpbkwke.

Abbreviation

CCM, CTT coupled motif

Circadian clocks

Nothing entraps man more than the conscious observation of the passing of time. We currently live within a culture where diverse aspects of our existence are tracked, traced, and documented by smart devices, often to the minute. However, long before the advent of sophisticated timekeeping devices, human behavior, and life on earth still tracked daily and seasonal oscillations in day length through intrinsic biological clocks. Although innately aware of the passing of time through daily sleep–wake cycles and seasonal variations in environmental light and temperature, the first record of individuals recognizing rhythmic daily oscillations in organism behavior in other species occurred in the 4^th century B.C. when Androsthenes, an admiral under Alexander the Great, described the daily rhythm of leaf movement of the tamarind trees in the Persian Gulf [1]. Despite these early observations, it was not until 1729 that de Mairan, a French astronomer, published the rhythmic pattern of daily leaf movements of heliotropic plants and their persistence away from external time cues in total darkness [2]. These early studies laid a foundation for behavioral research focusing on the rhythmic movements and biological processes across species that were aligned with our daily light–dark cycles. A central focus was placed on whether they were innate biological clocks that could persist in the absence of environmental cues.

These studies were extended to animal research in the late 19^th and early 20^th centuries when daily rhythms in animals were formally documented through reports of pigment rhythms in arthropods and behavioral patterns in rats [3, 4]. Similar daily behavioral oscillations were quickly discovered in diverse animals including honeybees [5, 6, 7], fruit flies [8, 9], dinoflagellates [10], and humans [11], leading to codification of the term circadian [12]. Today, we describe any process that satisfies three core rules as a circadian process: (1) It must involve ~24‐h cycles that are self‐sustaining in the absence of any environmental cues. (2) It must be entrainable and able to be resynchronized to environmental variables such as light. (3) It must be temperature compensated and persist with a 24‐h rhythm across a wide range of temperatures. Despite breakthroughs in behavioral studies, how organisms maintained a 24‐h rhythm in the absence of any environmental cues remained an enigma until the advent of genetic approaches in the second half of the 20^th century.

Genetic revolution

Over the past 50 years, our understanding of circadian rhythms has exploded due to our ability to define the molecular components that maintain and entrain diverse aspects of organism physiology to the 24‐h day–night cycle. We now know that over 40% of the human genome is circadian regulated [13]. In eukaryotes, we typically define the central timekeeping device (oscillator) as being composed of a transcription–translation feedback loop (TTFL). At their most simplistic level, these TTFLs are defined by a positive element that often involves transcription factors that drive the expression of clock‐controlled genes. A subset of these genes defines a negative element that, upon accumulation, repress circadian gene expression to close the TTFL. Tightly regulated post‐translational modifications modulate the TTFL to ensure a robust ~24‐h cycle. Although these central oscillators are sufficient to maintain a circadian rhythm, we now know that there are diverse peripheral feedback loops that couple almost all aspects of physiology to the core circadian clock, leading to diverse impacts of circadian dysregulation on human health [14, 15]. As a result, there have been significant efforts to target circadian clock components as treatments for diseases ranging from anti‐aging, cancer, diabetes, bipolar disorder, and obesity, among many others [16, 17, 18, 19, 20]. Thus, there is keen interest in understanding how clock components are dynamically regulated to mediate selective protein–protein interactions both within the core clock and between key regulators in peripheral feedback loops.

PAS/CRY proteins as mediators of protein–protein interactions and environmental sensing

Comparative analysis of the molecular components governing the TTFL in diverse eukaryotes identifies core conserved features that facilitate the use of model organisms to inform on the mammalian clock (Fig. 1). A striking feature in these systems is the recurrent use of period‐ARNT‐single‐minded (PAS), light‐oxygen‐voltage (LOV), and cryptochrome (CRY) proteins within the TTFL (Table 1). These domains and proteins function as environmental sensors and molecular scaffolds, dictating the rhythmic formation of protein complexes essential to the maintenance of circadian rhythms and facilitating the integration of light, redox state, and other environmental variables into the TTFL. Their central role in mediating crosstalk between the TTFL and disease‐relevant pathways has recently identified these PAS and CRY proteins as drug targets (Table 2), in part due to the presence of solvent‐filled pockets within their core protein scaffolds that mediate protein–protein interactions and have their ancestral origin as cofactor‐binding sites in photoactive members of the family (Fig. 2) [21, 22, 23, 24, 25, 26].

Fig. 1.

Comparative transcription‐translation feedback loops of fungal, insect, and mammalian circadian clocks. (A) Fungal system. The LOV domain protein white collar‐1 (WC1) and PAS domain protein white collar‐2 (WC2) form the white collar complex (WCC), which binds the clock‐box promoter in the dark to activate transcription of frq. Upon light activation of WC1 LOV domain, additional WC1 monomers are recruited, enabling WCC to bind the proximal light‐response element (PRLE) increasing transcription of frq and promotes transcription of vvd and circadian clock genes (ccgs). The FREQUENCY (FRQ) protein assembles with FREQUENCY‐interacting RNA helicase (FRH) and casein kinase‐1 (CK1) to form the FCC complex, which together with VVD, feeds back to inhibit WCC activity. (B) Insect system, PAS transcription factors, Clock and Cycle, bind to the E‐Box promoter to activate transcription of per and tim. Light‐activated cryptochrome (CRY) promotes Timeless (TIM) degradation via Jetlag (JET), destabilizing PER and contributing to feedback repression. (C) Mammalian system. CLOCK and BMAL1 bind to the E‐box promoter to activate per and cry transcription. PER:CRY complexes feedback to inhibit CLOCK:BMAL1 activity. CK1δ/ε phosphorylates PER, leading to destabilization, while FBXL3/21 promotes CRY degradation, furthering the TTFL.

Table 1.

PAS, LOV, and CRY proteins in model organism circadian clocks.

Organism	Protein	Domain class	Clock role	Interaction partners	Citation
Drosophila melanogaster	CLOCK (dCLK)	PAS (PAS‐A/B)	Positive arm (activator)	CYCLE (CYC), PER/TIM	[166, 167, 168]
Drosophila melanogaster	CYCLE (CYC)	PAS (PAS‐A/B)	Positive arm (activator)	dCLK; PER/TIM	[167, 168]
Drosophila melanogaster	CRY (dCRY)	CRY	Environmental sensor (light)	TIM, JET, FBXL proteins	[108, 134]
Mus musculus	CLOCK	PAS (PAS‐A/B)	Positive arm (activator)	BMAL1, NPAS2	[169, 170]
Mus musculus	BMAL1 (ARNTL)	PAS (PAS‐A/B)	Positive arm (activator)	CLOCK/NPAS2; coactivators (p300/CBP); PER/CRY	[170, 171]
Mus musculus	CRY1 / CRY2	CRY	Repressive arm	PER1/2; FBXL3/FBXL21; CLOCK–BMAL1	[172, 173]
Neurospora crassa	WC‐1	LOV–PAS	Positive arm and light sensor	WC‐2, FRQ/FRH, VVD	[174, 175, 176]
Neurospora crassa	WC‐2	PAS (PAS‐A/B)	Positive arm (activator)	WC‐1; FRQ/FRH	[176]
Neurospora crassa	VIVID (VVD)	LOV	Environmental sensor (photoadaptation)	WC‐1/WCC; FRH	[47, 177]
Arabidopsis thaliana	CRY1 / CRY2	CRY	Environmental sensor (blue light)	CIBs/SPA–COP1 axis	[178, 179]
Arabidopsis thaliana	ZTL (ZEITLUPE)	LOV–PAS (F‐box)	Environmental sensor/modulator (blue light)	GI; TOC1/PRR5	[180, 181]
Arabidopsis thaliana	FKF1	LOV–PAS (F‐box)	Environmental sensor/photoperiod (blue light)	GI; CDFs	[182]

Table 2.

Small‐molecule ligands of circadian and hypoxia PAS/CRY proteins.

Compound	Target	Mechanism	Binding site	Disease/use	Development stage	Citation
Belzutifan (MK‐6482, PT2977, Welireg)	HIF‐2α (EPAS1)	Inhibitor—disrupts HIF‐2α/ARNT dimer (allosteric)	HIF‐2α PAS‐B pocket	VHL syndrome tumors, Renal cell carcinoma	Approved 2021 (USA); ongoing Phase III trials	[21]
PT2385	HIF‐2α	Inhibitor—disrupts HIF‐2α/ARNT dimer (allosteric)	HIF‐2α PAS‐B pocket	Renal cell carcinoma	Phase I/II completed (predecessor to belzutifan)	[183]
Casdatifan (AB521)	HIF‐2α	Inhibitor—disrupts HIF‐2 dimer (allosteric)	HIF‐2α PAS‐B pocket	Renal cell carcinoma	Phase I ongoing (promising efficacy)	[184, 185]
NKT‐2152	HIF‐2α	Inhibitor—disrupts HIF‐2 dimer (allosteric)	HIF‐2α PAS‐B pocket	Renal cell carcinoma	Phase I/II ongoing	[184]
DFF332	HIF‐2α	Inhibitor—disrupts HIF‐2 dimer (allosteric)	HIF‐2α PAS‐B pocket	Renal cell carcinoma	Phase I completed; development halted	[184, 186]
CLK8	CLOCK protein	Inhibitor—disrupts CLOCK:BMAL1 heterodimer; blocks CLOCK nuclear entry	CLOCK PAS‐B domain	Circadian dysregulation (research tool)	Preclinical (probe compound)	[25]
CCM (Core Circadian Modulator)	BMAL1 protein	Modulator—binds BMAL1 PAS‐B, alters conformation	BMAL1 PAS‐B cavity	Inflammation, metabolic disease (theoretical)	Preclinical (probe compound)	[26]
KL001	CRY1/2	Stabilizer—prevents FBXL3‐mediated degradation; lengthens period	FAD pocket	Metabolic disease (preclinical); circadian probe	Preclinical	[156]
KL101	CRY1	Stabilizer—isoform‐selective CRY1 activator	FAD pocket	Circadian probe (CRY1‐specific)	Preclinical	[24]
KL201	CRY1	Stabilizer—isoform‐selective CRY1 activator	FAD pocket	Circadian probe (CRY1‐specific)	Preclinical	[157]
KL044	CRY1/2	Stabilizer—improved analog of KL001	FAD pocket	Circadian probe	Preclinical	[158]
TH301	CRY2	Stabilizer—selective CRY2 activator	FAD pocket	Circadian probe (CRY2‐specific)	Preclinical	[24]
TH303	CRY1	Stabilizer—lengthens period by inhibiting CRY1 degradation	FAD pocket	Circadian probe (CRY1‐specific)	Preclinical	[23]
TH129	CRY1	Stabilizer—lengthens period by inhibiting CRY1 degradation	FAD pocket	Circadian probe (CRY1‐specific)	Preclinical	[23]
GO044	CRY1/2	Stabilizer—shortens period	FAD pocket	Circadian probe	Preclinical	[159]
GO200	CRY1/2	Stabilizer—shortens period	FAD pocket	Circadian probe	Preclinical	[159]
SHP656	CRY2	Stabilizer—lengthens period	FAD pocket	Glioblastoma (preclinical); circadian probe (CRY2‐specific)	Preclinical	[162]
SHP1705	CRY2	Stabilizer—lengthens period; derivative of SHP656	FAD pocket	Glioblastoma (clinical candidate); circadian probe (CRY2‐specific)	Phase I completed	[187]
Compound 41	CRY1/2	Stabilizer—lengthens period	FAD pocket	Circadian probe	Preclinical	[160]
Compound 50	CRY2	Stabilizer—lengthens period	FAD pocket	Circadian probe (CRY2‐specific)	Preclinical	[161]
KS15	CRY1/2	Inhibitory—disrupts interactions between CRY and CLOCK:BMAL1 heterodimer	FAD pocket	Breast cancer; circadian probe	Preclinical	[22, 164]

Fig. 2.

(A) Neurospora crassa VIVID (NcVVD, PDBID: 2PDR). The protein adopts the conserved mixed αβ PAS/LOV fold, consisting of a five‐stranded antiparallel β‐sheet (Aβ, Bβ, Gβ, Hβ, and Iβ) flanked by α‐helices (Fα, Cα, Dα), forming a cleft that houses the FAD cofactor. (B) ARNT PAS‐B domain (PDBID: 3F1N) displays a similar conserved αβ PAS/LOV fold, with five‐stranded antiparallel β‐sheet with surrounding helices. (C) Example architecture of a CRY using Drosophila CRY (PDBID: 4GU5). The N‐terminal α/β domain (3–225, gray), C‐terminal α‐helical domain (226–518, wheat), serine loop (42–53, light blue), phosphate binding loop (PBL) (249–263, cyan), protrusion motif (PM) (288–306, pink), lid loop (420–466, purple), CC helix/$\alpha$22 (497–518, lime green), CTT (519–542, red), FAD (yellow). (D) Mammalian CRYs form an analogous fold, with the primary pocket (magenta) and secondary pocket (cyan) empty (shown using mCRY1: PDBID 5T5X).

This review will focus on dissecting the molecular mechanisms gating PAS and CRY signal transduction, with a specific emphasis on how photoactive members of the family inform on how cofactor or small‐molecule chemistry within the solvent‐filled pockets can dictate conformational changes important for signal transduction. For readers interested in the role of PAS domains in mammalian clocks [13], or the structural architecture and photochemistry of CRYs [27, 28], we note that there have been recent comprehensive reviews on these topics.

PAS/LOV domains

PAS domains, and their photoactive flavin‐binding LOV subclass, evolved from prokaryotic sensory systems, where they typically act as input modules in two‐component signaling pathways. In these prokaryotic systems, they often mediate adaptation to environmental variables such as oxygen, redox status, light, and the presence of metabolites or xenobiotics [29, 30]. They are now recognized as being present in all domains of life where, despite low levels of sequence conservation (as low as 20% identity), they retain a structural scaffold that enables PAS/LOV modules to integrate into multidomain signaling proteins to mediate both sensory responses and protein–protein interactions to regulate physiological responses [13, 29, 30]. Central to their function is a conserved mixed β‐fold, where a five‐stranded β‐sheet—denoted as Aβ, Bβ, Gβ, Hβ, and Iβ, with 2‐1‐5‐4‐3 topology (the conserved spatial arrangement of the antiparallel β‐strands)—serves as a central scaffold flanked by helical segments on each side (Fig. 2A,B). The β‐scaffold forms a cleft that is cradled by the α‐helical interface to form a solvent‐filled or hydrophobic pocket amenable to cofactor‐binding or small‐molecule recognition. The unique PAS/LOV architecture optimally couples diverse protein–protein interaction surfaces to the active site cleft, enabling coupling of cofactor chemistry or small‐molecule recognition to protein dynamics that can modulate protein–protein interactions within the circadian clock [13, 31, 32, 33, 34].

These protein–protein interaction surfaces can broadly be divided into four core regions: (1) The central β‐scaffold that often mediates PAS/LOV homo‐ and heterodimerization [35, 36, 37, 38, 39, 40, 41], (2) the α‐helical interface formed by Fα, a single long helix running along the β‐sheet, and three shorter helices (Cα, Dα, and Eα), which shape the conserved ligand‐binding cleft [21, 32, 40, 41, 42], (3) N‐terminal (A'α) or C‐terminal (Jα) helices peripheral to the PAS/LOV core domain that often mediate conformational changes in response to light, oxygen, or ligand binding [43, 44, 45, 46, 47], and (4) a disordered region connecting the C‐terminal β‐strands, Hβ and Iβ termed the HI‐loop (Fig. 2A,B) [13, 31, 32, 40]. While each of these regions plays important roles in mediating PAS/LOV signal transduction, we will particularly focus on the HI‐loop, which plays a unique role in mammalian circadian clock regulation and drug discovery.

HI‐loop as a locus for PAS protein–protein interactions and structural dynamics

The PAS ligand‐binding cavity, or pocket, can accommodate a variety of small‐molecule cofactors, including chromophores like FMN, FAD, or riboflavin, as well as oxygen sensing heme groups, carboxylic acids, and charged ions [41]. Despite the diversity of chemical cofactors, the PAS fold allows binding of these ligands without altering its basic architecture [48]. Interestingly, in many mammalian PAS proteins this ancestral pocket is solvent filled and is typically involved in protein–protein interactions. Central to these protein complexes is a disordered HI‐loop, where the three core mammalian clock proteins contain conserved tryptophan residues (PER2:Trp419, CLOCK:Trp362, BMAL1:Trp427), which are crucial for PAS‐mediated homo‐ and heterodimerization and competitive interactions between CLOCK, CRY, and histones to regulate rhythmic control of gene expression [13, 31, 32].

The conserved Trp residues within HI‐loops directly regulate four core protein–protein complexes within the mammalian circadian clock. First, in the PAS protein dPER, Trp482 in the PAS‐B β‐sheet surface (analogous to Trp448_mPER1, Trp419_mPER2, and Trp359_mPER3) stabilizes the protein–protein interface by inserting into a hydrophobic pocket of the PAS‐A domain of the dimerizing protein, where Trp mutations disrupt homodimerization (Fig. 3A) [34].

Fig. 3.

PAS domains stabilize protein:protein interactions (A) Homodimer of dPER (PDBID: 3RTY). Trp482 (yellow) of the HI‐loop inserts into the hydrophobic pocket of PAS‐A of the dimerizing protein, stabilizing the protein–protein interface. (B) CLOCK:BMAL1 heterodimer (CLOCK in pink, BMAL1 in blue, PDBID: 4F3L). Trp427_BMAL1 of the PAS‐B HI‐loop inserts into the pocket of CLOCK PAS‐B, interacting with Trp284_CLOCK of the AB loop. (C) Heterodimer of the isolated PAS‐B domains of HIF‐2α and ARNT (HIF‐2α in blue, ARNT in pink, PDBID: 3F1N). A solvent‐filled cleft within HIF‐2α PAS‐B is filled with ethylene glycol (yellow and red sticks). The isolated PAS‐B domains form an alternative dimer than that which is present in larger complexes (shown in 3D). (D) Heterodimer of ARNT and HIF‐2α with belzutifan bound to PAS‐B domain of HIF‐2α (HIF‐2α in blue, ARNT in pink, PDBID: 7W80). Belzutifan (yellow) bound to the pocket of HIF‐2α, forcing displacement of Met252_HIF‐2a from the solvent‐filled pocket, disrupting dimerization.

Second, examination of the CLOCK‐BMAL1 heterodimeric complex reveals that the HI‐loop and solvent exposed cleft of the PAS‐B domains are integral to PAS‐B dimerization [32]. The CLOCK‐BMAL1 PAS‐B domains make extensive contacts between the HI strands of BMAL1 and both the helical interface (Eα and Fα) and a loop connecting the Aβ and Iβ strands (AB loop) (Fig. 3B). Central to PAS‐B dimerization is insertion of Trp427 of the BMAL1 PAS‐B domain into the solvent‐filled cleft of CLOCK, which is formed by the helical interface, the AB loop, and the HI strands of the β‐scaffold. While an analogous Trp residue in the CLOCK PAS‐B HI‐loop is not involved in PAS dimerization, the residue plays an intricate role in regulating protein–protein interactions and circadian regulation in two additional protein complexes. Namely, Trp362 on CLOCK PAS‐B inserts itself into a solvent exposed cleft of CRY to dictate CRY‐mediated repression of circadian transcription [31]. Further, the CLOCK PAS‐B HI loop, including Trp362, forms extensive contacts with histones to alter E‐box nucleosome structure [40]. In these manners, the HI‐loop and solvent‐filled ancestral ligand‐binding pockets function as hotspots linking protein dynamics to regulation of protein complexes important for circadian function. Notably, these allosteric hot spots are not unique to circadian clocks but are broadly shared in PAS domain containing transcription factors relevant to disease.

The importance of interactions between the HI‐loop and the solvent‐filled cleft on the helical interface mirror regulatory interactions between the heterodimeric PAS transcription factors hypoxia inducible factor (HIF) and ARNT, which have been extensively targeted in drug discovery as a treatment for von Hippel–Lindau disease associated renal cell carcinoma [21, 42]. HIF‐ARNT forms a regulatory complex that is structurally homologous to CLOCK‐BMAL1, where solvent‐filled cavities are present within their PAS‐B domains (Fig. 3C,D). While HIF‐ARNT does not contain Trp residues analogous to those in CLOCK‐BMAL1 HI‐loops, the PAS‐B domains form similar dimeric contacts important for HIF‐ARNT regulation. Particularly, the ARNT PAS‐B HI strands and HI‐loop contact Fα at the helical interface of HIF‐2α and the AB loop to stabilize HIF‐ARNT PAS‐B dimerization (Fig. 3D) [21, 42]. Early identification of these solvent‐filled cavities by Kevin Gardner begged the question of what the structural effects of designing small molecules that target the large HIF cavity and smaller ARNT cavity would be [49, 50, 51, 52].

Detailed structural studies indicated that these cavities are coupled to dynamics within the β‐scaffold and associated HI‐loop that directly impact the ability of the PAS‐B domains to form heterodimers [39, 49, 50, 51, 53]. For instance, NMR studies of ARNT revealed a fragile native state conformation that could interconvert between two structural forms via a 3‐amino acid register shift of the Iβ strand that disrupted ARNT PAS‐B dimerization [54, 55]. Studies of a Y456T variant under both normal and high‐pressure NMR studies indicated that interconversion between the two folds was coupled to the solvent‐filled active site and involved cis‐trans isomerization of a proline residue (Pro449) in the HI‐loop, thereby directly coupling the solvent‐filled cleft with HI‐loop dynamics and PAS dimerization [54, 56].

Drug discovery in PAS proteins

The role of the solvent‐filled cleft in regulating HI‐loop dynamics as well as protein–protein interactions identified PAS proteins as viable targets in drug discovery. Early efforts focused on the larger cavity in HIF, which revealed impacts on PAS dimerization that have direct impacts in drug discovery. High‐throughput fragment‐based screens identified putative ligands that could bind to the HIF‐2α cavity and disrupt HIF‐ARNT PAS‐B heterodimerization [49, 52]. These initial studies led to the development of belzutifan [57], a HIF‐2α inhibitor approved for malignancies related to von Hippel–Lindau disease [21, 58]. Structural studies reveal that belzutifan binds to the solvent‐filled cleft of HIF‐2α, altering the conformation of Met252 within the AB loop and pushing it closer to the ARNT PAS‐B HI‐loop and disrupting dimerization (Fig. 3D). Solution biophysics experiments of HIF‐2α agonists also implicated movements of Met252 and His293 (Fα helix contacting solvent pocket) toward the ARNT HI‐loop in allosteric mechanisms regulating function [58]. Combined, these studies demonstrated that the solvent‐filled clefts are intricately involved in regulating conformational dynamics at the PAS‐B dimer interface, whereby they can be targeted by small molecules to impact physiological functions as therapeutic treatments. Analogous approaches have been employed to target the PAS‐B domain of the aryl hydrocarbon receptor (AhR) as FDA approved treatments for psoriasis and atopic dermatitis [59]. Despite these successes, limited attempts have been made to target the ancestral ligand‐binding pockets of CLOCK and BMAL1.

Computational screens of the PAS domains of CLOCK have covered ~2 million compounds, finding 100 potential hits of candidate molecules, one being CLK8 which is predicted to bind to the PAS‐A pocket of CLOCK, disrupting CLOCK‐BMAL1 heterodimerization [25]. Cell‐based assays showed CLK8 indeed interfered with CLOCK:BMAL1 heterodimerization, resulting in enhanced circadian rhythm amplitude, while the period remained unchanged. Crystallographic and biophysical drug discovery efforts have identified a synthetic ligand, core circadian modulator, that inserts into a pocket of the PAS‐B domain of BMAL1. Core circadian modulator produces dose‐dependent changes in PER2 oscillations and notably suppresses inflammatory and phagocytic pathways in macrophages [26]. A key limitation for the rational design of small molecules regulating CLOCK/BMAL1 activity is a lack of understanding of how cofactor chemistry or small‐molecule binding to the solvent‐filled pockets dictates protein dynamics to selectively disrupt contacts at the PAS‐B interfaces. Currently, our best understanding of PAS allostery stems from the study of the light‐regulated members of the PAS superfamily within the LOV subclass that are directly involved in circadian rhythms in plants and fungi.

LOV domains as a model of PAS signal transduction

LOV domains form a subclass of the PAS superfamily that originally broadly referred to PAS proteins containing cofactors sensitive to light, oxygen, or voltage (Fig. 4A,B) [29]. Recently, the term LOV domain more narrowly describes PAS proteins that bind a flavin cofactor and undergo blue light‐driven formation of a flavin‐cysteinyl C4a adduct through a conserved Cys residue in a GRNCRFLQ motif, which facilitates organism adaptation to changes in environmental light intensity and spectral qualities [60, 61, 62]. LOV domains can largely be divided into two subclasses: (1) Modular LOV proteins, where the LOV domain is coupled to diverse sensory outputs to impart blue light‐regulated control of cellular processes, and (2) Short LOV domains (sLOV) that lack accessory domains and typically function as light‐driven protein–protein interaction modules to regulate target proteins. Their ability to couple photon absorption to reorganization of protein–protein interaction motifs within PAS domains has made them an ideal model for PAS allostery and as sensors in optogenetic tools [61].

Fig. 4.

Signal transduction in LOV proteins. (A, B) Structures of LOV domains VVD (A, PDBID 2PDR) and AsLOV2 (B, PDBID 4HHD) that contain a PAS core (light blue) and either Ncap (yellow) or Ccaps (salmon) that are involved in signal transduction. (C) Signaling mechanism in LOV1 domains (CrLOV1: PDBIDs 8QI9 and 8QIW) adduct formation in the light‐state (orange) results in minimal movement of residues at the reface (Leu101, Val103). Signaling stems from Gln120 rotation altering contacts with Thr21 within Bβ. A tilt of the flavin upon adduct formation leads to subtle movements of H‐bonding residues (Asn99) to impact dynamics at the GH‐loop and possible salt bridges linking the GH‐loop to Iβ (Arg91‐Asp122). (D) LOV2 signaling (AtLOV2, PDBID: 6S46) involves global changes. Adduct formation in the light‐state (salmon) induces rotation of Phe470 and Gln489. Rotation of Gln489 alters contacts with Asn390 in Bβ. (E) VVD (light: yellow, PDBID: 2PDR; dark C71S mimic: light gray, PDBID: 2PD8) signaling stems from N5 protonation and rotation of Gln182 that alters contacts with Ala72. H‐bond changes induce rotation of Cys71 from a buried (aligning with Ser71 mutant) to an exposed position to H‐bond with the Asp68 side chain. (F) ZTL signaling diverges from other LOV domains and is independent of the Gln residue (dark‐WT, cyan PDBID: 5SVG). A G46S light‐state mimic (light orange, PDBID: 6WLP) demonstrates light‐induced movement of Phe156 into the active site inducing reordering of the C terminus and formation of a salt bridge between Arg125 (GH‐loop) and Glu158 in a manner analogous to that observed in LOV1 proteins (Fig. 4C). Residues at the reface of FMN (Leu107, Leu109) do not undergo conformational changes.

Initial models of LOV allostery stemmed from structural and computational analysis of three model systems: the tandem‐LOV domain containing phototropins, which couple photoactivity of LOV1 and LOV2 domains to regulation of a C‐terminal histidine kinase domain to mediate phototropic responses in plants [45, 63, 64, 65, 66], the sLOV protein vivid (VVD) that modulates photoadaptation in fungal circadian clocks [46, 47, 67, 68, 69, 70], and the bacterial stress response sensor YtvA [38, 71, 72]. Early studies focused on the β‐scaffold as a nexus for signal transduction due to its role in mediating LOV‐LOV dimers that mirrored those observed in isolated PAS domains across the PAS superfamily (Fig. S1) [37, 38, 39, 53]. Notably, it was quickly recognized that signaling in LOV proteins is highly flexible, where even within closely related proteins the magnitude and direction of signal transduction can vary due to subtle differences in residue identity near the flavin‐binding pocket [36, 73, 74, 75].

For instance, despite retaining moderate sequence identity and photochemical mechanisms, the LOV1 and LOV2 domains of phototropins demonstrated significant differences in dynamic properties. Specifically, LOV1 proteins underwent only modest conformational changes, but LOV2 photochemistry was coupled to large‐scale conformational rearrangements that dictated histidine kinase activity [76, 77, 78]. Computational and mutational studies of the isolated LOV1 and LOV2 domains identified differences within the β‐scaffold at the reface of the flavin as differentiators in LOV‐mediated signal transduction. Specifically, Phe (LOV2)→Leu (LOV1) mutations of a residue within Hβ inverted protein dynamics in LOV2 proteins to impart LOV1‐type dynamics (Fig. 4C,D) [75]. Computational studies confirmed significant differences in photoinduced dynamic motions in these two domains, whereby adduct formation altered the β‐scaffold to dictate dynamics within two flexible loops Gβ‐Hβ (GH‐loop; LOV1) and the HI‐loop (LOV2) to differentiate LOV1/LOV2 signal transduction [76].

Subsequent detailed characterization of the isolated LOV2 domain of Avena sativa phototropin (AsLOV2) [44, 45, 79, 80] and the sLOV protein VVD [46, 47, 67, 68, 69, 70] altered our ideas of LOV signal transduction. In both systems, N‐terminal (A'α) and/or C‐terminal (Jα) extensions to the PAS core (Ncap/Ccap) were identified as being central to their regulation. In both cases, initial models of LOV allostery focused on photoinduced formation of the flavin‐cysteinyl adduct that induced sp³ hybridization of the C4a position and protonation of the N5 position on the flavin isoalloxazine ring [64, 65, 81, 82]. Crystal structures of both proteins identified that N5 protonation induces rotation of a ‘conserved’ Gln residue that altered H‐bonding interactions with residues in Bβ (Fig. 4C–E) [43, 44, 47, 68]. NMR, solution biophysics, and crystal structures of light‐state proteins indicated that these proximal changes in H‐bonding interactions propagated to A'α and/or Jα to induce local unfolding events that impacted intra‐ and inter‐protein interactions [44, 45, 47, 68]. Subsequent studies of LOV proteins lacking the photoactive Cys residue indicated that these proteins retain an ability to signal due to formation of the flavin neutral semiquinone (FH^•) that is also protonated at the N5 position [83]. As a result, a consensus model of LOV signal transduction developed whereby N5 protonation was both necessary and sufficient to induce global conformational changes and signal transduction by invoking rotation of the ‘conserved’ Gln residue [83]. These studies led to a LOV signal‐transduction model primarily focused on the singular signal‐transduction axis involving the β‐scaffold and associated Ncap/Ccap linked by N5 protonation and rotation of the ‘conserved’ Gln.

Subsequent studies of diverse systems indicated that while the above allosteric mechanisms are accurate, they do not adequately encompass the breadth of PAS/LOV allostery. Rather, alternative signaling modes exist to link cofactor chemistry to diverse protein–protein interaction surfaces. Recent identification of LOV proteins lacking the ‘conserved’ Gln residue has been identified in diverse systems, most prominently within the zeitlupe (ZTL) family of proteins that mediate light‐regulated degradation of circadian and photoperiodic clock components. Some members of this family contain Leu residues at the position occupied by the canonical Gln residue but retain photoactivity [84, 85]. Structural and biochemical studies of ZTL indicated the isolated LOV domain formed a parallel dimer mediated by the β‐scaffold in the dark, which upon light activation underwent a 180° degree rotation [36, 37, 86]. Detailed structural and computational studies identified movement of a conserved Phe within Iβ, sensitive to the electronic state of FMN as central to mediating signal transduction (Fig. 4F) [36, 86]. Notably, structural studies of bacterial LOV proteins had identified similar Phe motions as key mediators of conformational changes across the LOV family [87, 88, 89]. These modes of signal‐transduction mirror initial models of signaling in LOV1 domains, where Phe→Leu variants alter LOV protein dynamics. Alternative mechanisms of signaling were cemented by a tour‐de‐force study by Dietler et al., where they employed an optogenetic tool based on YtvA to assess the impact of Gln mutations in LOV signal transduction [90]. These studies indicated that in contrast to previous studies, and existing LOV signaling models, the Gln residue is not required for light‐regulated conformational changes and that alternative ‘Gln‐less’ signaling mechanisms were likely widespread.

Recent time‐resolved crystallography studies of the isolated LOV1 and LOV2 highlight the varied signal‐transduction landscape in LOV domains, where multiple allosteric trajectories can impact PAS/LOV signal transduction [91, 92]. Studies of the LOV2 domain of Arabidopsis thaliana phototropin‐2 (AtLOV2) indicate a complex global conformational response initiated by both reorganization within the β‐scaffold due to rotation of the conserved Gln residue [91]. Key to the global conformational response was the presence of the Phe residue within Hβ that undergoes a 90° rotation following photoexcitation (Fig. 4D). In addition, rotation of Gln489 alters H‐bonding contacts with Asn390 within Bβ to couple adduct formation to reorganization of Ncap and Ccap elements. These two signaling nexuses induce increased dynamics within both the GH‐ and HI‐loops to initiate a global conformational response [91]. These conformational changes contrast with time‐resolved studies of the LOV1 domain of Chlamydomonas reinhardtii phototropin‐1 (CrLOV1), which demonstrates a more muted conformational response consistent with previous MD‐studies [76, 92]. The LOV1 protein undergoes only modest conformational changes in residues near the FMN binding pocket, with increased dynamics within the GH‐loop coupled to reorientation of the FMN cofactor and rotation of the conserved Gln residue (Fig. 4C) [92]. These time‐resolved crystallographic studies provide detailed information regarding the allosteric trajectories and how LOV proteins employ diverse aspects of cofactor chemistry to elicit different conformational responses based on subtle differences in residue identity within the cofactor‐binding pocket.

The combined studies of LOV domains identify them as highly flexible signaling molecules capable of coupling cofactor chemistry to diverse signaling outputs, where the precise allosteric trajectory and magnitude of conformational responses could vary widely based on subtle differences in residue identity at key allosteric sites. This begs the question of what are the molecular determinants differentiating the direction and magnitude of signal‐transduction mechanisms in PAS/LOV proteins. Even in homologs between closely related species, subtle differences in residue identity can impart divergent modes of conformational dynamics or impart an ability to sense additional environmental variables such as osmotic stress, temperature, and cellular redox status [73, 74, 93]. Computational approaches that attempt to decipher the allosteric landscape in LOV proteins have identified that the flavin‐binding pocket (solvent‐filled cleft) couples cofactor chemistry, subtle H‐bonding changes, and electronic effects to global alteration in dynamic modes linking the β‐scaffold, flexible loops, and helical interface [76, 86, 94, 95, 96, 97, 98]. These dynamic modes are highly sensitive to the residue identity at key evolutionary hotspots. These models do not yet have the predictive power to identify how cofactor chemistry or small molecules will alter conformational dynamics and signaling in PAS/LOV domains from sequence and structure alone; however, combined computational and experimental approaches in LOV proteins afford promise to understand how artificial small molecules may predictively alter PAS signaling in disease relevant proteins and shed light on how similar small‐molecule scaffolds can elicit divergent effects (agonist, antagonist, etc.) in these drug targets by altering PAS protein dynamics in a predictive manner.

Cryptochromes

CRYs are conserved across all domains of life and bear homology to photolyases, which use an FAD cofactor and an antenna chromophore to catalyze light‐dependent DNA repair [27, 99, 100]. Most CRYs, however, have lost the ability to bind to or repair DNA. Instead, they play crucial roles in regulating circadian rhythms in plants and animals, in addition to numerous other functions regulating growth, development, and adaptation in diverse organisms [27, 28, 101, 102, 103, 104]. CRYs have also been implicated in magnetoreception across various species (insects, birds, plants), but this role remains highly controversial [28, 105, 106].

CRYs involved in circadian regulation can be broadly divided into two categories: type I (light‐sensitive) and type II (light‐insensitive) [107]. Despite sharing a common photolyase‐homologous core, these two types have evolved distinct mechanisms according to their roles. Type I CRYs (exemplified by Drosophila melanogaster CRY, or dCRY) function as blue‐light photoreceptors that directly respond to environmental light, whereas type II CRYs (such as mammalian CRY1 and CRY2) act as light‐independent transcriptional repressors within the circadian clock feedback loop [28, 108, 109, 110]. An additional subtype (type IV) of animal CRYs has been implicated in magnetosensing, but are not believed to be involved in circadian rhythms and thus will not be a major focus of this review [111, 112]. Comparative analysis of the photoactive dCRY and the nonphotoactive mammalian CRY1/2 has illuminated how subtle structural adaptations (differences in key signaling surfaces, residue interactions, and regulatory loops/tails) underlie the loss of photoreception in vertebrates and the gain of protein–protein interaction capacity. These differences are not only fundamental to CRY biology but also present unique opportunities for pharmacological targeting of CRY1 and CRY2.

Structural composition

All CRYs are structurally defined by a photolyase homology region (PHR) and a variable C‐terminal tail (CTT) that imparts species and isoform specific function [27, 28, 113]. While structural information regarding the CTT is limited to a single structure with an intact CTT [114, 115, 116], the PHR is well defined and conserved across all CRY/photolyase family members. The PHR is composed of two subdomains: an N‐terminal α/β domain and a C‐terminal α‐helical domain, both of which contain key protein–protein interaction motifs important for regulation of circadian rhythms and CRY‐mediated signal transduction (Fig. 2C). Central to CRY function are two solvent exposed pockets involved in signaling and protein interactions. These are known as the primary pocket and the secondary pocket based on their ancestral roles in FAD binding and DNA repair (primary pocket) and antennae pigment binding (secondary pocket) in photolyases [99] (Fig. 2D). As discussed below, the functional role and ability to bind photoactive pigments of these pockets has diverged in type I and type II CRYs. However, key structural features within the primary and secondary pockets essential for regulation of protein–protein interactions have been retained. Specifically, both pockets are defined by flexible loop regions, whose dynamics gate both cofactor chemistry and recognition of protein–protein interaction partners as well as dictate isoform specific functions in type II CRYs.

The primary pocket is formed by three flexible loop regions named based on their ancestral roles in regulating photolyases or the photoactive type I CRY. These include the phosphate binding loop (PBL) that facilitates recognition of the DNA phosphodiester backbone (photolyases) [117] or accommodates the CTT (dCRY) [116], a ‘lid’ loop, which interacts with and secures the CTT in the primary pocket of dCRY [116], and a protrusion motif (PM) that expands the size of a cleft within the protein surface that recognizes DNA (photolyases) [117] or the CTT (type I CRYs) [116]. As discussed below, dynamics within these flexible loop motifs couple cofactor chemistry within the primary pocket to regulation of protein–protein interactions, as well as facilitate the recognition of small‐molecule regulators of CRY function.

The secondary pocket is formed by a flexible surface loop termed the serine‐rich loop, which contributes to isoform specific differences in mammalian CRYs [118]. While the secondary pocket binds an antenna pigment in photolyases and some CRYs, this pocket is empty in both photoactive type I and type IV CRYs and is currently of unknown function. In type II CRYs, this pocket is repurposed to bind protein partners instead of a chromophore. These functional differences in type I and type II CRYs are discussed in detail below.

Type I vs type II cryptochromes: FAD Binding and signaling interactions

The evolutionary transition from type I to type II CRYs was accompanied by subtle amino acid substitutions and structural reorganization that resulted in the loss of direct photoreception and an enhanced capacity for protein–protein interactions. In broad terms, type II CRYs have repurposed the chromophore‐binding sites of photolyases/type I CRYs as protein interaction surfaces. The primary and secondary pockets that in photoactive CRYs bind FAD or an antenna pigment are utilized in mammalian CRYs to recruit key partners (FBXL3, CLOCK, etc.) rather than chromophores. Below, we compare how specific structural elements differ between the light‐sensitive dCRY and the light‐insensitive mammalian CRY1 and CRY2 isoforms, highlighting the consequences for signaling and interactions.

Primary pocket

In type I CRYs, the primary pocket binds an FAD cofactor and serves an autoinhibitory role in the dark. The CRY C‐terminal tail (CTT) is tucked into this pocket, blocking it until light activation undocks the CTT [119]. Notably, the phenylalanine‐rich tip of the dCRY tail occupies the same location in the pocket that a DNA lesion would occupy in photolyase [115, 116]. This mimicry underscores how dCRY uses its tail to ‘plug’ the active site. Upon light‐induced formation of an anionic semiquinone (FAD^•–) and the consequent release of the tail, the pocket becomes available for binding to other proteins. Specifically, the N terminus of TIM can bind to the primary pocket of light‐activated dCRY, wrapping around the PHR domain in a manner strikingly similar to how DNA wraps around photolyase (Fig. 5A) [120]. This light‐dependent exposure of the primary pocket is the switch that enables downstream signaling in type I CRYs.

Fig. 5.

(A) dCRY:TIM in the light state (PDBID: 8DD7). The N terminus of TIM (blue) binds the primary pocket of dCRY (gray) next to FAD (yellow). (B) The secondary pocket (orange) of mCRY1 (gray) (mCRY1:mPER2, PDBID: 4CT0) binds the HI‐loop (red) of the PAS‐B domain (purple) of CLOCK (lime green) (PDBID: 4F3L). The transactivation domain (TAD) of BMAL1 (not shown) competes with mPER2 (cyan) to bind the mCRY1 CC/α22 helix (magenta).

Several specific residues in dCRY's primary pocket region govern this light‐dependent switch and interaction with TIM. For example, two conserved His and Arg237 in the pocket have been identified as important for TIM recognition and for facilitating CTT release [121]. Within the CTT itself, Glu530 and Ser526 are essential for CRY–TIM binding (in Drosophila) and for analogous CRY–PER interactions in other species [122]. Furthermore, residues in the vicinity of the pocket contribute to the photochemical reaction efficiency: Met331 and Cys337 (located in the so‐called ‘sulfur loop’ of the PHR), together with Cys523 (in the tail‐to‐PHR connector loop), appear to gate electron transport along the Trp tetrad and influence the lifetime of the FAD^•– radical in dCRY (Fig. S2) [115]. In summary, the primary pocket of type I CRYs is a multifunctional site which binds FAD, holds the inhibitory tail, and, upon light activation, engages target proteins like TIM.

By contrast, type II CRYs (mammalian CRY1/2) lack a stably bound FAD in the primary pocket. Purified CRY1 and CRY2 are generally FAD‐empty and have greatly reduced affinity for FAD compared to insect CRYs. Several amino acid differences explain this loss of flavin binding [123]. In mouse CRY1, residues that correspond to FAD‐contacting positions in dCRY are altered or reoriented in a way that occludes the pocket. For example, the side chains of His355 and Gln289 in mCRY1 are rotated into the space that FAD occupies in dCRY, and His355 packs against Arg358 [115] (analogous to Arg381 in dCRY, which normally forms a stabilizing salt bridge with Asp410 to secure FAD [124]). Additionally, two residues that in dCRY make contacts with the FAD phosphates, Ser265 and Arg237, are replaced by Gly250 and His224 in mCRY1, eliminating key interactions that would stabilize cofactor binding [115]. The net result of these substitutions is that the primary pocket of mammalian CRYs prefers to remain ligand‐free and structurally stable without FAD.

Freed from the need to bind a chromophore, type II CRYs have co‐opted the primary pocket for protein binding. In particular, the F‐box protein FBXL3, which ubiquitinates CRY for degradation, docks into the primary pocket of CRY1/2 in order to recognize and bind its target (Fig. S3) [125]. Structural and mutational analyses have shown that two cysteine residues in the CRY1 pocket (Cys412 and Cys414) are critical for FBXL3 engagement [126]. Thus, what was a FAD‐binding cleft in light‐activated CRYs has become a degron‐binding interface in light‐independent CRYs in an evolutionary repurposing of the pocket.

Secondary pocket

The secondary pocket also diverged in function between type I and II CRYs. A major difference lies in the flexibility and conformation of the loop that caps this pocket. In mammalian CRY1/2 (type II), the loop, spanning roughly residues 35–53, forms part of the secondary pocket's upper rim and tends to adopt a structured, open conformation, which leaves the pocket accessible for protein interactions [31, 127]. In dCRY (type I), by contrast, the homologous loop contains a bulky tryptophan (a residue that is conserved in insect CRYs but not in mammals) that protrudes into and fills the secondary pocket, effectively blocking access. In mammalian CRY1/2, this position is a glycine, creating a vacant pocket [127]. These sequence differences mean that type I CRYs have a largely occluded secondary pocket, whereas type II CRYs have an open pocket surface available for binding partner proteins.

Consistent with this, mammalian CRYs use the secondary pocket as a docking site for transcription factors. The secondary pocket in CRY1/2 is known to bind the PAS‐B domain of the CLOCK protein through conserved Trp362 [31] (Fig. 5B). Mutational studies swapping key residues between insect and mammalian cryptochromes support the importance of this region. For instance, introducing Drosophila‐like residues into CRY1 at positions that differ in the loop/pocket (e.g. P39G, F14S, G106W, which make CRY1 more ‘Type I‐like’) was found to weaken CRY1's interaction with CLOCK:BMAL1 [127]. Conversely, Drosophila CRYs engineered to have a glycine instead of the native tryptophan in that loop could hypothetically gain interaction capacity. Thus, subtle sequence differences in the secondary pocket loop translate into functional divergence: type II CRYs retain an accessible pocket for binding CLOCK, whereas type I CRYs have an occluded pocket of unknown function.

CTT (C‐terminal tail)

Although type II CRYs are not light‐activated, their disordered CTTs still perform an autoinhibitory regulatory function analogous to that of the light‐released tail in type I CRYs. In mammalian CRY1, a specific segment of the CTT (encoded by exon 11 in humans) has been shown to modulate interactions with CLOCK. Structural modeling suggests that this exon 11 segment of hCRY1 can fold back onto the secondary pocket of the PHR domain, partially blocking CLOCK:BMAL1 from binding [128]. Indeed, a human CRY1 variant lacking exon 11 (referred to as CRY1Δ11) binds more tightly to CLOCK and acts as a stronger repressor, lengthening the circadian period. The absence of this autoinhibitory tail fragment is the molecular basis for a form of human delayed sleep phase disorder (DSPS) in which individuals carrying the CRY1Δ11 mutation have abnormally long circadian cycles [128, 129]. This finding highlights that, similar to the light‐regulated release of the CTT in dCRY, mammalian CRY1 keeps its tail in place to restrain activity until other signals intervene. Notably, CRY2 lacks an equivalent exon 11 segment, which may contribute to differences between CRY1 and CRY2 as discussed later [129].

Another parallel between insect and mammalian CRYs is observed in how the CRY CTT competes with partner proteins for binding. When the repressor protein PER2 associates with CRY1, it wraps around the CRY PHR domain in a manner remarkably similar to TIM binding to dCRY, effectively displacing the CRY1 tail from the pocket. Cryo‐EM structures show the PER2 globular CRY‐binding domain contacting the same face of CRY1 that the tail normally occupies [130]. This suggests that PER2 and the CRY1 tail have overlapping binding sites, and the tail must move (or be ejected) for PER to bind tightly. As evidence, certain mutations in the CRY tail can weaken PER binding. For instance, in mouse CRY2, Arg501 and Lys503 in the CTT are important for PER2 binding and for CRY2's transcriptional repression activity [131]. Modifications to the CRY1 tail can also affect PER binding: an interesting redox‐dependent mechanism exists whereby a disulfide bond can form between Cys412 and Cys363 in CRY1's PHR domain, and this intramolecular bond weakens the CRY1‐PER2 interaction. Tetrahedral coordination of a zinc ion (by Cys414 on the C‐terminal lid of mCRY1, His473 on α22 of mCRY1, and Cys1210 and Cys1213 on the C‐terminal of mPER2) favors reduction of this disulfide, thereby strengthening mCRY1‐mPER2 binding (Fig. S4) [126]. In summary, the CTT in type II CRYs dynamically modulates access of repressive partners (PER, CLOCK/BMAL1) to the PHR, a concept analogous to the light‐controlled tail dynamics in type I CRYs but achieved through different biochemical cues.

Finally, the CTT of mammalian CRYs also mediates interactions with the transcriptional activators of the clock. The transactivation domain (TAD) of BMAL1 has been shown to bind a C‐terminal helix (α22) preceding the CTT of CRY1/2 [132] (Fig. 5B). Key basic residues in the CRY1 tail (Arg483, Lys485, and Arg478) were found to contact the BMAL1 TAD in structural studies [133]. Thus, the CRY tail is a multifaceted interaction hub: It can occlude binding sites in an autoinhibitory fashion, and it can also directly bind transcription factors (BMAL1) or facilitate repressor recruitment (PER) when appropriately modified or displaced. These mechanisms underscore a recurring theme: whether by light (in insects) or by other regulatory inputs (in mammals), CRY CTTs act as switches that control the assembly of larger circadian complexes. Notably, the CTT interacts with the flexible PBL, lid‐motif, and an empty primary pocket to dynamically regulate protein–protein interactions, thereby providing a putative small‐molecule binding site capable of regulating both CRY stability and signal transduction in manners analogous to photoactive CRYs. For these reasons, photoactive CRYs became early models of how cofactor chemistry may modulate the CTT and flexible loop dynamics to modulate protein–protein interactions.

Signal transduction in photoactive CRYs

Our best understanding of how small‐molecule or cofactor chemistry may alter conformations within the PBL and flexible lid loops (collectively referred to as the CTT coupled motif or CCM) to propagate changes in protein dynamics to the CTT and protein–protein interaction surfaces stems from photoactive members of the CRY family. To date, the two best characterized systems are the circadian clock photoreceptor dCRY, which is the only member of the CRY family to be structurally characterized with an intact CTT, and an animal‐like CRY from Chlamydomonas reinhardtii (aCRY). Although signal‐transduction mechanisms in these proteins remain hotly debated, they provide keen insight into CRY regulation.

In these systems, conformational changes coupled to cofactor chemistry can be isolated to proximal responses within the FAD cofactor‐binding pocket, and distal responses within the CCM, CTT and protein–protein interaction surfaces. Two key conserved surfaces are of note: 1) A surface cleft lined by two conserved His/Asn and two conserved Trp residues that form essential contacts with DNA photolesions in DNA photolyase, the CTT of dCRY, and key contacts involved in isoform‐selective drug design in mammalian CRYs (Fig. S5A), and 2) A terminal helix (α22) directly preceding the CTT involved in protein–protein interactions with BMAL1 important for transcriptional repression (Fig. 5B) [128]. We briefly summarize these allosteric trajectories linking FAD chemistry to regulation of signal transduction with possible relevance to mammalian systems of interest to drug discovery (discussed below).

dCRY

Light‐induced structural rearrangements in dCRY are essential for TIM binding and light‐induced degradation of dCRY [27, 108, 120, 121, 134]. In the dark, the CTT of dCRY is tightly docked into the primary pocket and cradled by the CCM lined by the conserved His/Asn and Trp residues (Fig. 2C and Fig. S5A). CTT binding is facilitated by a network of hydrophobic and π–π stacking interactions (notably a conserved FFW motif at Phe534/Phe535/Trp536) (Fig. 6A) and by hydrogen bonds between the CTT (Glu530, Ser526) and the PHR core (Trp422, Trp314, Tyr158, Gln159) [115]. Upon absorption of blue light, the FAD cofactor is photoexcited (FAD*). The excited FAD* then accepts an electron via a chain of conserved tryptophans called the Trp tetrad (W420→W397→W342→W394), generating an anionic semiquinone radical FAD^•– [101, 135]. Photoreduction of the FAD results in proximal changes in H‐bonding interactions. Specifically, a crucial internal salt bridge (Arg381‐Asp410) is disrupted and a stabilizing cation–π interaction between His378 and Phe534 is weakened, collectively destabilizing the CTT's attachment to the pocket (Fig. 6A) [120, 124]. Additionally, computational approaches and site‐directed mutagenesis suggest that electron transfer leads to protonation of His378, altering the hydrogen bond network and further loosening the CTT [120, 121, 136, 137]. Consequently, the CTT is released from the primary pocket. This exposes a binding surface on CRY, allowing the N terminus of TIM to insert into the pocket vacated by the CTT.

Fig. 6.

Signaling mechanisms in photoactive CRYs. (A) Photoactivation of dCRY is believed to disrupt a conserved salt bridge above photoactive FAD as observed in CRY‐Tim complexes (cyan, PDBID: 8DD7). An additional signaling pathway has been proposed whereby protonation of His378 alters H‐bonding contacts at the FFW motif of the CTT and adjacent loop motifs. (B) In aCRY (dark‐state, green, PDBID: 6FN3; light‐state, blue, PDBID: 8Z3X) light‐induced electron transfer across the Trp triad ends in formation of a TyrO^• radical via protonation of Asp321. Asp321 rotates away from α22 altering stabilizing interactions with Arg492 and possible unfolding of the α22 helix. (C) The aCRY signal‐transduction mechanism operates through an alternative face as dCRY signaling. dCRY signals through the DNA‐binding cleft and the conserved His/Trp residue interface (orange), which interacts with the CTT (pink). aCRY signaling is initiated at the alternative face involving the Trp triad and Tyr373 to impact α22 H‐bonds with a disordered strand involving Asp321 and Asp323. Both signaling mechanisms impact the CCM and α22 to couple to conformational changes within the CTT.

Notably, key elements of this allosteric pathway linking FAD photochemistry to structural changes at the distal CCM and CTT remain controversial or poorly understood. First, the role of His378 protonation in CTT release has been questioned by some limited proteolysis and time‐resolved SAXS experiments [138, 139]. However, His378 is highly conserved in the CRY/photolyase family and undergoes light‐dependent protonation in photolyases where the protonated His is essential for DNA repair. These degrees of conservation and positioning in the distal protein–protein interaction surface strongly suggest a role in regulating CRY signal transduction. Further, limited proteolysis experiments indicate that the PBL undergoes light‐dependent ordering with concomitant disordering of both the CTT and portions of α22, leading to altered conformational dynamics within the CCM [119]. Notably, similar CCM dynamics are intricately involved in DNA lesion recognition in photolyases [140], and regulation of photoactive type IV CRYs [112], albeit often with divergent directions and magnitudes. Currently, the factors dictating CCM dynamics remain underexplored due to the difficulty of extracting detailed structural information in transiently structured protein regions involved in signal transduction.

aCRY

Chlamydomonas aCRY is part of a unique clade of CRY proteins that are bifunctional, retaining photolyase activity [141, 142], while also being directly involved in light‐regulated control of the circadian and sexual life cycles [143, 144]. In aCRY, the ground state of the FAD cofactor is believed to be the one‐electron reduced neutral hydroquinone radical (FADH^•) where blue‐light excitation drives electron transfer via a Trp electron transfer chain that is terminated by a Tyr residue, leading to formation of the flavin anionic hydroquinone (FADH⁻) and a long‐lived deprotonated Tyr radical (TyrO^•) [145, 146, 147]. Time‐resolved IR and visible spectroscopy demonstrate that TyrO^• formation initiates a global conformational change involving reorientation of an extended ~100 residue amino acid containing CTT. Hydrogen‐deuterium‐exchange mass spectrometry (HDX), small‐angle‐Xray scattering (SAXS), and single‐molecule FRET demonstrate that reorientation of the CTT is coupled to movement of the α22 helix [148, 149]. While these biophysical measurements provided a global context for how FAD chemistry may dictate conformational changes within animal‐like CRYs, they lack atomic level details outlining an allosteric trajectory from the FAD active site to protein–protein interaction surfaces.

Recent time‐resolved serial femtosecond crystallography (Tr‐SFX) studies of aCRY have provided keen insight into how local changes in cofactor chemistry can dictate allosteric changes in distal protein–protein interaction surfaces in CRY proteins [150]. These studies differentiate proximal light responses within the FAD‐binding pocket, and a putative mechanism of signal transduction inducing large‐scale conformational changes within the α22 helix. Central to both aCRY photochemistry and signal transduction is the presence of an Asn residue directly adjacent to the N5 position of the isoalloxazine ring of FAD (Fig. S5B). The presence of an Asn (Asn395) at this position favors stable formation of FADH^• and the ability to generate an FADH⁻ signaling state with high quantum yield [150]. Moreover, time‐resolved studies demonstrate that initial photo‐excitation and formation of FADH^• is achieved through an unusual proton transfer event from His309 to the N5 position of the flavin isoalloxazine ring via formation of an Asn395 imidic acid leading to reorientation of both the His309 and Asn395 side chains (Fig. S5B) [150]. Coupled to FADH^• was initial formation of the Tyr cationic radical, which is deprotonated by on a nanosecond time scale to form TyrO^• and protonated Asp321 [147, 150]. Protonation of Asp321 led to rotation of Asp321 away from the α22 helix altering stabilizing H‐bonding interactions and initiating local disordering of α22 (Fig. 6B,C) [150]. Although these studies were conducted in a construct lacking the CTT and studied formation of the FADH^• ground state, they provide keen insight into how distal chemistry can alter key structural elements important for signal transduction across the CRY superfamily. Further, they highlight flexibility in signal‐transduction mechanisms within the CRY family. Namely, whereas dCRY signal transduction is believed to proceed through changes in H‐bonding interactions through the conserved DNA‐binding cleft (Figs 5A, 6C) and CCM ultimately affecting α22 and the CTT due to formation of FAD^•– [27], in aCRY signal transduction is initiated by TyrO^• and proceeds through an alternative face of the CRY protein to ultimately impact the analogous α22 and CTT protein–protein interaction surfaces (Fig. 6C) [150].

Combined, studies of photoactive CRYs demonstrate how the CRY/photolyase structure facilitates coupling of ancestral cofactor‐binding pockets to impact diverse protein–protein interactions by modulating the dynamics of flexible motifs. Given that these same flexible motifs dictate CTT dynamics and disease‐relevant protein–protein interactions in mammalian CRYs, it begged the question of whether small molecules could be designed to target the empty primary and secondary pockets of mammalian CRYs as possible therapeutics.

CRYs as drug targets

CRY1 vs CRY2—Small‐molecule selectivity

Given the central role of CRYs in circadian regulation and their impact on downstream physiological processes, CRY1 and CRY2 have emerged as promising targets for therapeutic intervention. Disrupted circadian rhythms are implicated in a wide range of pathologies including cancer [151], metabolic syndrome [152], and neurodegenerative diseases [153]. Consequently, considerable effort has been directed toward developing small‐molecule CRY modulators, with the goal of selectively stabilizing or destabilizing a specific CRY isoform to correct clock dysfunction or exploit circadian vulnerabilities in disease [154]. An important consideration in these efforts is the inherent structural differences between CRY1 and CRY2, which medicinal chemists can leverage to achieve isoform selectivity.

Structurally, CRY1 and CRY2 are highly similar overall, especially in the conserved PHR domain and primary pocket. However, they differ significantly in their CTTs and in the conformational dynamics of a key ‘gatekeeper’ Trp within the primary pocket. In CRY1, this gatekeeper residue (Trp399) predominantly adopts an ‘open’ or Trp‐out conformation, pointing away from the pocket cavity. In CRY2, the corresponding Trp417 more often assumes a Trp‐in conformation, tucked inward and partially occupying the pocket (Fig. 7). These preferred conformations are stabilized by structural differences in the lid loop: in CRY1, Gln407 in the lid loop helps keep Trp399 flipped out, whereas in CRY2, a Phe424 structurally occupies the analogous position, favoring Trp417 packing inside the pocket. The result is a subtly different pocket shape and volume between CRY1 and CRY2. Such differences, while minor from a protein‐fold perspective, can have large effects on ligand binding, and indeed are thought to underlie why some compounds preferentially bind one isoform over the other [155]. In addition, the divergent CTT sequences between CRY1 and CRY2 may modulate how the pocket behaves allosterically or how a ligand might stabilize the protein.

Fig. 7.

KL101 binds to the intrinsic ‘in’ conformation of mCRY1's (pink, 6KX6) gatekeeper tryptophan (W399) (pink, PDBID: 6KX6). TH301 binds to the intrinsic ‘out’ conformation of mCRY2's (blue, 6KX8) gatekeeper tryptophan (W417) (blue, PDBID: 6KX8).

The first known small molecule to target CRYs was KL001, discovered in a chemical screen for clock modulators. KL001 binds to the primary pocket and stabilizes both CRY1 and CRY2 by blocking the interaction of CRY with FBXL3. By preventing CRY degradation, KL001 lengthens the circadian period and dampens the amplitude of rhythms [156]. However, because KL001 has equal affinity for CRY1 and CRY2, substantial efforts were made to derive analogs with improved isoform selectivity. This led to a series of second‐generation compounds: KL101 (a CRY1‐selective stabilizer) [24], KL201 (a CRY1‐selective stabilizer with a distinct chemical scaffold) [157], TH303 and TH129, CRY1‐selective stabilizers [23], and TH301, a CRY2‐selective stabilizer [24]. These molecules all bind in or near the primary pocket of CRY, yet discriminate between the two isoforms, producing either CRY1‐specific or CRY2‐specific effects on circadian period.

The mechanisms underlying isoform‐selective binding of these compounds are dictated by the structural differences discussed above. In general, a compound may selectively bind to the isoform whose primary pocket conformation (especially the gatekeeper Trp and lid loop arrangement) matches the compound's preferred binding mode. For example, KL001 is relatively agnostic to the Trp‐in vs Trp‐out difference in that it can accommodate either conformation, consistent with its ability to bind both CRY1 and CRY2 with equal affinity. By contrast, KL101 and KL201 have higher affinity for CRY1 because they are optimized to fit the intrinsic Trp‐out conformation of the CRY1 pocket and TH301 and SHP656 have higher affinity for CRY2 because they are optimized to fit the intrinsic Trp‐in conformation of the CRY2 pocket. TH301 is unusual in that it can induce a conformational change: It binds to CRY1 in a way that pushes Trp399 into an inward orientation (Fig. 7). Forcing the CRY1 Trp into the pocket makes CRY1 more similar to CRY2 from the perspective of the ligand. Accordingly, when Trp399 is artificially constrained in an ‘in’ state, the CRY1 affinity for CRY2‐selective compounds increases and vice versa for CRY2. Experiments swapping portions of the CRY1 and CRY2 lid loops provided further evidence: altering the lid loop residues that stabilize the Trp orientation was sufficient to reverse the isoform selectivity of certain compounds [155].

The CTT also influences small‐molecule selectivity. Swapping the CTTs of CRY1 and CRY2 was found to swap their sensitivities to isoform‐selective ligands like KL101 and TH301. A CRY1 protein carrying the CRY2 tail loses its usual response to CRY1‐specific compounds, and vice versa [155]. Further, truncation studies on CRY2 showed that the unique exon 10 segment of the CRY2 tail is required for compounds like TH301 to discriminate CRY2 from CRY1. Exon 10 in CRY2 is proposed to fold back near the primary pocket [24], potentially contacting bound ligands or altering the shape of the pocket allosterically. These findings reinforce that the tail and pocket form an integrated unit when it comes to ligand binding: differences in the tail between CRY1 and CRY2 translate to differences in how compounds bind and stabilize the PHR.

Beyond KL001 and its immediate derivatives, a variety of other CRY‐targeting compounds have been identified. For example, KL044 is a more potent analog of KL001 [158]. GO044 and GO200 are compounds discovered in a cell‐based screen that, interestingly, shorten the circadian period [159], suggesting that they might destabilize CRY or act via a different mechanism than the period‐lengthening stabilizers. Additional molecules include those known by their screening identifiers: compound 41 [160] and compound 50 [161], both of which showed efficacy in cellular models of metabolic disease, and SHP656, a synthetic agonist that emerged from a chemical optimization and was shown to bind CRY2 [162]. Interestingly, virtually all CRY‐activating (stabilizing) compounds tested to date cause circadian period lengthening, regardless of whether they are CRY1‐selective or CRY2‐selective. Compounds like GO200 that produce period shortening likely do so by a different mode.

From a therapeutic standpoint, the burgeoning arsenal of CRY modulators holds promise for a wide array of conditions. For metabolic disorders, maintaining robust CRY activity can be beneficial. For instance, compounds 41 and 50 (CRY stabilizers) have been shown to improve glucose tolerance and prevent hyperglycemia in diabetic mouse models [160, 161]. In cancer, where circadian disruption is often a hallmark, CRY‐targeting drugs may impair cancer cell survival. The compound SHP656, which stabilizes CRY, was found to inhibit the growth of glioblastoma cells in vitro and significantly extend survival in mice bearing glioblastoma tumors [162]. There is also evidence that altering CRY levels can influence disease outcomes: overexpression of CRY1 in obese, insulin‐resistant mice improved insulin sensitivity and lowered blood glucose [18, 163], suggesting that increased CRY1 activity has antidiabetic effects. Conversely, inhibiting CRY function has been explored for anticancer therapy. For example, a small‐molecule CRY1/2 inhibitor dubbed KS15 was reported to dampen CRY activity and showed potential in suppressing breast cancer cell growth [22, 164], as well as reducing tumor size in a mouse model [165]. These studies illustrate that downstream pathways can be impacted by tuning the circadian clock through modulating CRY activity.

Despite significant advances in CRY drug discovery, several challenges remain. For one, the secondary pocket of CRY remains an almost completely unexplored site for pharmacological intervention. Given its critical role in mediating interactions with CLOCK, BMAL1, and PER in the circadian complex, this pocket presents an attractive target for future drug design. A compound that binds in the secondary pocket could, for example, prevent CLOCK:BMAL1 from docking to CRY or disrupt a necessary PER–CRY interface, thereby modulating the clock in a different manner than primary pocket stabilizers. Such compounds have the potential to achieve greater isoform specificity by taking advantage of differences in the secondary pocket and surrounding loops between CRY1 and CRY2 or could be used in combination with primary pocket binders to fine‐tune CRY activity.

Secondly, the precise mechanisms of action for some clock‐modulating compounds are still not fully understood. The period‐shortening molecule GO200 and the CRY inhibitor KS15 are two examples where the biochemical targets and effects on CRY structure are not clearly defined. Additionally, new compounds such as TH303 and TH129 have exhibited CTT‐independent activity [23]. This suggests that these molecules may bind in alternative sites or orientations, potentially by rearranging the lid loop or gatekeeper Trp. Such alternative binding modes could present both an opportunity, allowing for targeting CRYs in new ways, and a challenge, due to off‐target effects or allosteric changes. Understanding exactly how these compounds engage CRY will be crucial for further optimization.

Thirdly, isoform selectivity is presumed to be dictated by differences in conformational dynamics in the flexible PBL and lid loops, and the CTT. How small‐molecule chemistry at the primary binding pocket is coupled with conformational dynamics within these transiently structured motifs is still poorly understood, but essential to the design of improved drugs. While our understanding of factors gating conformational changes within these regions has been improved by photoactive members of the CRY family, these studies are still limited in scope. Additional studies of both photoactive and nonphotoactive CRYs from diverse systems are needed to provide a keen understanding of how subtle sequence differences and small‐molecule chemistry can impact conformational dynamics in CRY proteins.

Perspective

Over the past 20 years, advances in structural and computational approaches have provided deep insight into the molecular mechanisms of PAS/CRY proteins within circadian clocks. These studies demonstrate that the PAS/CRY scaffolds are evolutionarily designed to couple cofactor chemistry within ancestral ligand‐binding sites to regulate complex protein–protein interactions to mediate dynamic control of environmental variables and stable circadian rhythms. Central to our understanding of these systems is detailed characterization of photoactive members of the family that afford keen insight into how cofactor chemistry alters PAS/LOV/CRY allostery and protein dynamics. Despite significant progress, our ability to rationally design small‐molecule therapeutics targeting PAS/CRY proteins remains limited. These limitations are largely due to difficulties in understanding how subtle differences in residue identity can have dramatic impacts on transiently structured loop motifs and signaling interfaces that impart isoform and species‐specific differences in both signal transduction and protein dynamics. Future studies employing new approaches to experimentally study these transiently structured motifs to link experimental and computational approaches are needed. Further, augmenting these studies by focusing on closely related, but functionally distinct PAS/CRY proteins promises to expand our predictive power in designing small‐molecule regulators of PAS/CRY proteins as new therapeutics.

Author contributions

EDB, AEL, and BDZ drafted the manuscript.

Supporting information

Fig. S1. Representative LOV dimers.

Fig. S2. Residues in the primary pocket of dCRY contribute to photochemical reaction efficiency.

Fig. S3. W428 in the C‐terminal of FBXL3 (magenta) binds the primary pocket of mCRY2 (grey) (PDBID 4I6J).

Fig. S4. mPER2 (cyan) wraps around mCRY1 (grey) (PDBID 4CT0).

Fig. S5. Signal transduction in photoactive animal CRYs.