Unwinding the differences of the mammalian PERIOD clock proteins from crystal structure to cellular function

Abstract

The three PERIOD homologues mPER1, mPER2, and mPER3 constitute central components of the mammalian circadian clock. They contain two PAS (PER-ARNT-SIM) domains (PAS-A and PAS-B), which mediate homo- and heterodimeric mPER-mPER interactions as well as interactions with transcription factors and kinases. Here we present crystal structures of PAS domain fragments of mPER1 and mPER3 and compare them with the previously reported mPER2 structure. The structures reveal homodimers, which are mediated by interactions of the PAS-B β-sheet surface including a highly conserved tryptophan (Trp448_mPER1, Trp419_mPER2, Trp359_mPER3). mPER1 homodimers are additionally stabilized by interactions between the PAS-A domains and mPER3 homodimers by an N-terminal region including a predicted helix-loop-helix motive. We have verified the existence of these homodimer interfaces in solution and inside cells using analytical gel filtration and luciferase complementation assays and quantified their contributions to homodimer stability by analytical ultracentrifugation. We also show by fluorescence recovery after photobleaching analyses that destabilization of the PAS-B/tryptophan dimer interface leads to a faster mobility of mPER2 containing complexes in human U2OS cells. Our study reveals structural and quantitative differences between the homodimeric interactions of the three mouse PERIOD homologues, which are likely to contribute to their distinct clock functions.

In mammalians many physiological, behavioral, and biochemical processes are regulated in a day-time dependent (circadian) manner. The approximately 24 h period is generated by a circadian clock, which is operated by molecular feedback loops. In the main feedback loop, the bHLH (basic-helix-loop-helix)-PAS (PER-ARNT-SIM) transcription factors mBMAL1/2, mCLOCK, and NPAS2 activate the expression of three PERIOD proteins (mPER1, mPER2, and mPER3) as well as two cryptochromes, mCRY1 and mCRY2. The mPER and mCRY proteins inhibit their own transcription, completing the circle of negative feedback. The daily regulated expression of mBMAL1 is ensured by a stabilizing feedback loop, in which the heme binding nuclear receptor REV-ERBα/β represses mBMAL1 expression, whereas ROR-α/β activates it (1, 2). Recently, nontranscriptional circadian oscillations of peroxiredoxin oxidation-reduction, hemoglobin dimer-tetramer transitions, and NADH/NADPH oscillations have been described in human red blood cells, which are interconnected with the transcriptional feedback loops in nucleated cells (3, 4).

The PERIOD proteins and the transcription factors mBMAL1/2, mCLOCK, and NPAS2 contain two tandemly organized PAS domains (PAS-A and PAS-B). The PAS domains mediate homo- and heterodimeric interactions between the mPER homologues (5–8) as well as interactions of the mPERs with mBMAL1/2, mCLOCK, and NPAS2 (9–12). These interactions regulate the stability and cellular localization of the mPERs and modulate the activity of the mBMAL1/mCLOCK transcription factor complex. Additionally, the PAS domains of NPAS2, mCLOCK, and mPER2 have been reported to bind heme (13–16). In the circadian clock, mPER1 and mPER2 proteins are found in large protein complexes, likely establishing multiple interactions via their PAS domains, the central CKIε/δ binding domain and the C-terminal mCRY binding region (17–19).

Studies with mPER knockout mice showed that mPER1 and mPER2 are more essential for circadian rhythmicity than mPER3 (20–22). mPER2 appears to positively regulate the expression of clock genes (mper1, mper2, mcry1, mbmal1) in vivo (20, 21), possibly by interacting with REV-ERBα and thereby modulating its effect on mBMAL1 transcription (23). In contrast, mPER1 knockout leads to decreased peak amounts of mPER2 and mCRY proteins in the nucleus, suggesting that mPER1 regulates their stability and/or nuclear entry through protein-protein interactions (20). The subtle effect of mPER3 deficiency on circadian behavior implies that mPER3 is more important for output functions. Indeed, mPER3 interacts with the nuclear receptor PPAR-γ via an N-terminal region including both PAS domains and a preceding predicted helix-loop-helix motive (24). This interaction represses the PPAR-γ activity and thereby inhibits the adipogenesis of mesenchymal stem cells. The importance of the per3 gene is also shown by its implication in the delayed sleep phase syndrome and the morning or evening sleep timing preferences observed in human populations (25, 26).

To provide insights into the molecular mechanisms underlying the distinct functions and molecular interactions of the three mPER homologues, we have solved crystal structures of the PAS domain regions of mPER1 and mPER3 and compared them with our mPER2 structure (27). In addition to a conserved PAS-B/tryptophan dimer interface, mPER1- and mPER3 homodimers are stabilized by interfaces located within the PAS-A domain (mPER1) or a predicted helix-loop-helix region N-terminal to the PAS-A domain (mPER3). These additional interfaces lead to increased affinities compared to the mPER2 PAS domain homodimers. We also provide evidence that the PAS domains of mPER1 and mPER3 might be able to bind heme. Furthermore, we show by luciferase complementation assays that the mPER homodimers observed in our crystal structures are also formed in HEK293 cells. Moreover, our fluorescence recovery after photobleaching (FRAP) analyses reveal that destabilization of the PAS-B/tryptophan dimer interface generates faster moving mPER2 containing complexes in human U2OS cells.

Results

Crystal Structures of Mouse PERIOD1 and Mouse PERIOD3.

We have determined crystal structures of the fragments mPER1[191–502] and mPER3[108–411], which include the two PAS domains (PAS-A and PAS-B), the αE helix C-terminal to PAS-B and about 25 residues N-terminal to PAS-A (Fig. 1, Fig. S1, and Table S1).

Fig. 1.

Crystal structures of mPER1[191–502] and mPER3[108–411] (A) Ribbon presentation of the mPER1[191–502] homodimer with molecule 1 shown in cyan, molecule 2 in yellow. The conserved Trp448 is shown as atomic stick figure. (B) Close-up view of the mPER1 PAS-A/αC dimer interface formed by antiparallel packing of the αC helices. The 2F_o-F_c electron density is shown in gray (1σ level). (C) Close-up view of the mPER1 PAS-A/N-terminal cap interface. (D) Ribbon presentation of the mPER3[108–411] homodimer with molecule 1 shown in magenta, molecule 2 in green. The conserved Trp359 is shown as atomic stick figure. (E) Close-up view of the mPER3 PAS-A/αC dimer interface. The composite omit map is shown in blue (2F_o-F_c, σ = 1). (F) Close-up view of the mPER3 PAS-A/N-terminal cap interface. His124 (αN) forms main- and side chain contacts to Thr131, Leu250, and Val 251 and water mediated H-bonds to Arg227 and Glu252. Lys127 (αN) forms a salt bridge to Glu252. Leu120 makes van der Waals contacts to Trp190 and Leu248. (A)–(F) Interacting residues and water molecules are highlighted as atomic stick figures and spheres. Numbers indicate distances in Å.

Both structures revealed noncrystallographic homodimers (Fig. 1 A and D). Each monomer contains two canonical PAS domains with a five-stranded antiparallel β-sheet (βA-βE) covered on one face by α-helices (αA, αA*, αB, αC) (Fig. 1 A and D, Fig. S1, and Table S2). The mPER homodimers are stabilized by interactions between the antiparallel PAS-B β-sheet surfaces. Central to this interface are the conserved tryptophans Trp448_mPER1 and Trp359_mPER3 [corresponding to Trp419_mPER2; (27)] as well as two phenylalanines, Phe444_mPER1/Phe355_mPER3 (= Phe415_mPER2) and Phe454_mPER1/Phe365_mPER3( = Phe425_mPER2) (Fig. 1 A and D, Figs. S2A and S3C).

The mPER2 homodimer also involved interactions of the PAS-A domain with helix αE and the PAS-B domain of the dimerizing molecule (27) (Fig. S2B). These interactions are not observed in the mPER1 and mPER3 structures because the relative orientation of the monomers is changed in these two homologues (Fig. S3A; Table S2). Instead, a second homodimer interface is observed in the PAS-A domains of mPER1 and mPER3, which is mostly mediated by contacts between their antiparallel αC helices (Fig. 1 B and E, Fig. S2B). Central to this interface is a tyrosine residue, Tyr267_mPER1 and Tyr179_mPER3. In mPER1, the presence of two glycine residues (Gly264 and Gly268) allows for a close approach of the two αC helices (Fig. 1B). In mPER3, the αC helices do not approach each other as closely as in mPER1, likely due to the exchange of Gly264_mPER1 to a bulky arginine (Arg176_mPER3) (Fig. 1E, Fig. S2B).

In all three mPER homologues the region N-terminal to the PAS-A domain (referred to as N-terminal cap) interacts with the PAS-A domain surface intramolecularly. Whereas the N-terminal cap of mPER2 is unstructured, it folds into a long α-helix (αN) and a β-strand (βN) in mPER1 and a shorter α-helix (αN) in mPER3 (Fig. 1 A and C, D and F, Fig. S3 A and B). Our mPER structures revealed a conserved interaction between Tyr200_mPER1/Tyr171_mPER2/Tyr112_mPER3 (N-terminal cap) and the PAS-A residue Trp278_mPER1/Trp249_mPER2/Trp190_mPER3 (Fig. 1 C and F, Fig. S3B, and ref. 27). In mPER1, the PAS-A residue Tyr233 constitutes a central residue of this interface, which contacts Leu202, Leu205, and Thr209 of the N-terminal cap (Fig. 1C). Additionally, Leu205 and Ile208 (αN) form van der Waals contacts to Trp278 and Leu338 (PAS-A). Thr209 and Thr213 (αN) hydrogen bond to Thr219, Ser221, and Gln237 of PAS-A. In mPER3, Tyr233_mPER1 is replaced by a histidine (His145_mPER3), which does not establish strong contacts to the N-terminal cap (shortest distance to Thr118: 4.5 Å). Instead, the N-terminal cap of mPER3 contacts the PAS-A surface via His124, Leu120, and Lys127 as detailed in Fig. 1F.

In the PAS-A βE-strand of mPER2 an LxxLL coactivator motive (³⁰⁶LCC³⁰⁹LL) has been identified, which plays a role in the interaction of mPER2 with REV-ERBα and possibly other nuclear receptors (23). Although the sequences are changed to ³³⁵PCC³³⁸LL in mPER1 and ²⁴⁵PCC²⁴⁸LT in mPER3, the coactivator motive regions (βE strands) of mPER1, 2, and 3 superimpose well (Fig. S3 A and B). However, Leu338_mPER1 and Leu248_mPER3 are buried in a hydrophobic pocket formed by Trp278_mPER1/Trp190_mPER3 (PAS-A) and Leu205_mPER1/Leu120_mPER3 of the N-terminal cap (Fig. 1 C and F, Fig. S3B). In mPER2, Leu309_mPER2 is less covered and the ßD-ßE loop preceding the coactivator motive is less ordered than in mPER1 and mPER3 (Fig. S3 A and B).

The αE helix of mPER1, mPER2, and mPER3 contains a functional nuclear export signal (NES; Fig. S1), which corresponds well to the consensus sequence L-x(2,3)-[LIVFM]-x(2,3)-L-x-[LI] (28). Whereas Leu489_mPER1/Leu460_mPER2/Leu399_mPER3, Ile493_mPER1/Ile464_mPER2/Ile403_mPER3, and Leu496_mPER1/Leu467_mPER2/Leu406_mPER3 pack against the αC’ helix of PAS-B, the C-terminal residues Leu498_mPER1/Met469_mPER2/Leu408_mPER3 point to the molecule surface (Fig. S4).

Analysis of mPER PAS Domain Interactions in Solution.

In order to prove the existence of the mPER1 and mPER3 homodimers in solution and to compare their affinities with the mPER2 PAS domain fragments (27), we have analyzed the fragments mPER1[197–502] and mPER3[108–411] by analytical gel filtration and analytical ultracentrifugation (Fig. S5). The dissociation constant of the mPER3[108–411] homodimer is 1.72 μM, which is comparable to the equivalent mPER2[170–473] fragment (K_D = 1.34 μM). For the mPER1[197–502] PAS domain fragment, however, we obtained a K_D of 0.15 μM corresponding to a 10 to 15 times higher affinity than mPER2 and mPER3 (Fig. S5G).

We have mutated the PAS-B interface residues Phe444, Trp448, and Leu456 of mPER1 as well as Trp359, Ile367, and Pro330 of mPER3 (Fig. S3C) to glutamate. In our gel filtration experiments, the mutations W359E and I367E totally and P330E partially disrupted the mPER3[108–411] homodimer (Fig. S5D). In the mPER1[197–502] fragment only the W448E mutation completely prevented homodimer formation. The F444E mutation was partially effective (Fig. S5 A and C) and the L456E mutation (corresponding to I367E in mPER3) totally ineffective. Analytical ultracentrifugation revealed that the mPER1[197–502]F444E mutant protein population, which elutes before the wild-type protein in our gel filtration experiments, corresponds to a homodimer with an enlarged hydrodynamic radius and not to a higher oligomeric state. We reasoned that in the dimeric F444E mutant population the two monomers are loosely held together by the remaining PAS-A/αC dimer interface. To prove this hypothesis we generated a F444E/Y267E double mutant in order to destabilize both, the PAS-B/β-sheet- and the PAS-A/αC dimer interface. As expected, the double mutant totally disrupted the homodimer (Fig. S5C). In contrast, the Y267E single mutation left the homodimer intact suggesting the PAS-B interface to be more important.

Sequence analysis predicts the existence of a helix-loop-helix motive N-terminal to the PAS-A domain of mPER1, 2, and 3 (5, 29, 30) and secondary structure predictions suggest this N-terminal region to be mostly α-helical. To explore its potential contribution to homodimer formation, we have determined the dimer affinity of the N-terminally extended mPER3 PAS domain fragment mPER3[32–411], which includes the predicted helix-loop-helix motive (Fig. S1B). Indeed, the K_D for the mPER3[32–411] homodimer was 0.4 μM corresponding to a roughly four times higher affinity compared to mPER3[108–411] (Fig. S5G). Moreover, none of our PAS-B dimer interface mutants disrupted the mPER3[32–411] homodimers (Fig. S5E). Furthermore, our CD spectra revealed an increased α-helical content of mPER3[32–411] compared to mPER3[108–411] (Fig. S6 B–D). Hence, our analysis suggests the presence of an additional α-helical homodimer interface between residues 32 and 107 of mPER3, which likely adopts a helix-loop-helix fold. Since mPER1 and mPER2 have also been predicted to contain a helix-loop-helix motive N-terminal to their PAS-A domain (Fig. S1B) (5, 29), we tested the effect of the W419E mutation, which is able to disrupt mPER2[170–473] and mPER2[128–473] PAS domain homodimers (27), in the N-terminally extended fragment mPER2[59–473]. Unlike mPER3[32–411], the mPER2[59–473] homodimers are efficiently disrupted by the W419E mutation (Fig. S5F).

Heme Binding of mPER1 and mPER3.

To find out, if mPER1 and mPER3, like mPER2 (14, 15), are able to bind heme in their PAS domains, we have incubated our purified mPER3[108–411], mPER3[32–411], and mPER1[197–502] proteins with heme, separated the heme exposed proteins via gel filtration chromatography and assessed heme binding by UV/VIS spectroscopy. For all three fragments we observed a shift of the absorption maximum from 390 nm (free heme) to about 420 nm (protein-bound heme, Fig. S7 A–C), suggesting that heme binds to our mPER1 and mPER3 PAS domain fragments.

PAS Domain Interactions are Essential for mPER Homodimerization in Mammalian Cells.

To investigate if the mPER crystal dimers are also present inside cells, we performed luciferase complementation experiments in HEK293 cells using wild-type and tryptophan mutant versions of the mPER proteins (Fig. 2). In our previously reported coimmunoprecipitation (Co-IP) experiments (27), mutation of the conserved Trp419_mPER2 (W419E) efficiently disrupted mPER2[128–473] PAS domain homodimers, but only had a subtle effect on the formation of full-length mPER2 homodimers in HEK293 cells. We concluded that the full-length mPER2 homodimer is additionally stabilized via non-PAS mPER2 regions, either directly (due to mPER2-mPER2 homodimer interactions) or indirectly by other interacting molecules such as the cryptochromes. To distinguish between these two possibilities, the luciferase complementation assay was performed with a C-terminally truncated mPER2[1–1127] fragment (mPER2ΔC), which is unable to interact with mCRY1 (Fig. 2 A and B). Compared to the wild-type mPER2ΔC fragment, the amount of homodimers (measured as luciferase activity) was reduced to about 40% when both monomers contained the W419E mutation and to about 75% when only one monomer was mutated (Fig. 2B). This result suggests that the PAS-B/tryptophan interface is the predominant homodimer interface of the full-length mPER2 protein and indirect mCRY-mediated interactions masked the W419E mutant effect in our Co-IP studies with full-length mPER2.

Fig. 2.

Tryptophan residues within mPER-mPER interfaces are critical for mPER homodimerization in mammalian cells. (A) Luciferase complementation assay—proof-of-concept: mPER2 and mCRY1 were expressed as fusion proteins with an N-terminal or C-terminal fragment of firefly luciferase in HEK293 cells. Strong and specific bioluminescence signals were detected upon mPER2 and mCRY1 fusion protein coexpression, but not if a truncated version of mPER2 with a deletion of the C-terminal mCRY binding domain—mPER2ΔC—or an irrelevant protein (βGAL) is coexpressed. Data are normalized to renilla luciferase activity, which was used as a transfection control. Shown are average values and error bars of two independent transfections. (B)–(D) mPER fragments were expressed as luciferase fusion proteins in HEK293 cells. The N-terminal luciferase fragment was fused to the N-terminus and the C-terminal luciferase fragment to the C-terminus of a corresponding mPER fragment. Shown are average values of four to five independent transfections with s.e.m. (*: p < 0.05; **: p < 0.005; ***: p < 0.001; n.s.: not significant; t-test). Each experiment was performed at least three times with similar results. (B) mPER2 without mCRY-binding domain (mPER2ΔC = mPER2[1–1127]) was expressed either as wild-type (wt) or as W419E mutant (mut). Mutation at position 419 in both binding partners (mut/mut) severely, mutation in one partner (wt/mut) moderately reduces bioluminescence signals. (C) mPER1[197–502] fragments were expressed either as wild-type (wt) or as W448E mutant (mut). Mutation at position 448 in both binding partners (mut/mut) severely, mutation in one partner (wt/mut) moderately reduces bioluminescence signals. (D) mPER3[108–411] fragments were expressed either as wild-type (wt) or as W359E mutant (mut). Mutation at position 359 in one or both of the binding partners (wt/mut and mut/mut) severely reduces bioluminescence signals.

In our luciferase complementation assay, homodimers of the mPER3[108–411] PAS domain fragment were largely disrupted by the W359E mutation (Fig. 2D), while the corresponding W448E mutation in mPER1 efficiently weakened dimerization of the mPER1[197–502] fragment (Fig. 2C). We conclude that the PAS-B/Trp homodimer interface of mPER1 and mPER3 is also present inside HEK293 cells. In agreement with our homodimer affinity measurements, about 90% of the mPER3[108–411] dimers are disrupted when Trp359 is mutated in one or both monomers (Fig. 2D), whereas the amount of mPER1[197–502] dimers is reduced to about 45% when both monomers contain the W448E mutation and to about 65% when only one monomer is mutated (Fig. 2C). Since we have identified an additional homodimer interface between residues 32 and 107 of mPER3, which prevents the W359E mutation from disrupting mPER3[32–411] homodimers in solution (Fig. S5E), the W359E mutation is unlikely to disrupt homodimers of full-length mPER3 in our luciferase complementation assay.

Role of Homodimers in the Formation of mPER Containing Complexes.

To analyze the potential role of mPER homodimers in the formation of mPER containing clock protein complexes, we have monitored their mobility in the cytoplasm of human U2OS cells by following Venus-fused mPER proteins in FRAP (fluorescence recovery after photobleaching) experiments (Fig. 3). While the fluorescence of tetrameric Venus-fused ß-galactosidase (MW = 580 kDa) recovered within a half-time (t_1/2) of 0.56 s, mPER1 and mPER2 complexes showed a slower recovery with t_1/2 = 0.76 s for mPER1 and t_1/2 = 1.73 s for mPER2. The lower mobility of the cytoplasmic mPER1 and mPER2 complexes suggests, that they are larger than tetrameric Venus-fused ß-galactosidase and possibly retained by interactions with other cytosolic proteins or structural elements. Mutation of the conserved Trp419_mPER2 (W419E) leads to a significantly faster fluorescence recovery of mPER2 containing complexes (t_1/2 = 1.06 s). The equivalent W448E mutation in mPER1 did not significantly affect the mobility of the mPER1 complexes, resulting in a wild-type like recovery rate with t_1/2 = 1.04.

Fig. 3.

mPER protein mobility in U2OS cells determined by fluorescence recovery after photobleaching (FRAP). (A) mPER proteins were stably expressed in U2OS cells as fusion proteins with the fluorescent Venus protein. They are localized in both cytoplasm and nucleus (depicted is mPER2). A representative bleach area of a FRAP experiment is indicated. The scale bar represents 10 μm. Lower shows a representative bleaching and recovery of fluorescent molecules in the bleach areas during the first two seconds for mPER2 wild-type (wt) or the mPER2 W419E mutant fusion proteins. (B) FRAP measurements: normalized average curves of fluorescence recovery in the cytoplasm expressing mPER2 wt (dark blue), mPER2 W419E (light blue), mPER1 wt (dark green), mPER1 W448E (light green), or βGAL (a similar sized control protein; black). Data from the first two seconds of recovery (boxed area) is enlarged. Red marks show t_1/2 for mPER2 wt and mPER2 W419E. (C) Comparison of half-time of recovery (t_1/2 ± sem; n: number of experiments with at least 10 cells per experiment): mPER2 wt (t_1/2 = 1.73 ± 0.13 s; n = 7) is significantly less mobile (p < 0.001, t-test) than the control protein βGAL (t_1/2 = 0.56 ± 0.02 s; n = 3). mPER2 W419E (t_1/2 = 1.06 ± 0.11 s; n = 3) displays a significantly increased mobility (p = 0.016, t-test) compared to mPER2 wt. In contrast, mPER1 mobility (t_1/2 = 0.76 ± 0.08 s; n = 4) is not increased upon mutation of W448E (t_1/2 = 1.04 ± 0.15 s; n = 4). There is a trend (p < 0.08, t-test), however, toward a lower mobility of mPER1 wt when compared to βGAL.

Discussion

To provide mechanistic insights into the nonredundant functions of the three PERIOD homologues mPER1, 2, and 3, we have determined crystal structures of homodimeric PAS domain fragments of mPER1 and mPER3 and compared them with the known crystal structure of mPER2 (27). While the PAS-B/tryptophan dimer interface is present in all three mPER homologues, the PAS-A-PAS-B/αE dimer interface of mPER2 is replaced by a PAS-A-PAS-A interface in mPER1 and mPER3, which is mediated by the two antiparallel αC helices (Fig. 1, Figs. S2 and S3). The different PAS-A dimer interactions result from the changed relative orientation of the two monomers in mPER1 and mPER3 (Fig. S3A), which is likely correlated with the nonconservative amino acid exchanges in the PAS-A-PAS-B/αE interface of mPER2 [(27) and Figs. S1B and S2B).

The center of the PAS-A/αC interface is formed by Tyr267 in mPER1 and Tyr179 in mPER3 (Fig. 1 B and E). Interestingly, this tyrosine is replaced by an alanine (Ala287_dPER) in the Drosophila PERIOD (dPER) homologue. This substitution enables the insertion of Trp482_dPER (corresponding to Trp448_mPER1/Trp419_mPER2/Trp359_mPER3) into the PAS-A domain binding pocket of the dimerizing molecule and hence the formation of a completely different dPER homodimer (27, 31). Gly264 and Gly268 of mPER1, which allow for a close approach of the dimerizing αC helices and hence the formation of a tighter PAS-A/αC interface than in mPER3, are conserved in mammalian PER1 homologues but not in other PER proteins or bHLH-PAS transcription factors. We propose, that the tight PAS-A/αC interface of mPER1 is responsible for the roughly 10 times higher affinity of mPER1 PAS domain homodimers (K_D = 0.15 μM) compared to mPER2[170–473] (K_D = 1.34 μM) and mPER3[108–411] (K_D = 1.72 μM) and accounts at least partly for the lower efficiency of PAS-B/tryptophan interface mutations in disrupting mPER1 homodimers in solution [Fig. S5 A and C and ref. 27) and in human HEK293 cells (Fig. 2 C and D; Fig. 3C).

Although mPER proteins do not have known sensory functions, mPER2 has been reported to bind heme as a cofactor in its PAS domains and in its C-terminal region (13, 14). Our UV/VIS spectroscopic analyses (Fig. S7 A–C) suggest that the PAS domains of mPER1 and mPER3 might also be able to bind heme. Cys215, which has been proposed as an axial ligand for heme binding to the PAS-A domain of mPER2 (13), is conserved and ordered in mPER1 (Cys244, αA*) but changed to a serine (Ser156) in mPER3 (Fig. S7D). Cys270, another potential heme ligand of mPER2 (13), is conserved in mPER1 and mPER3 (Cys299_mPER1, Cys210_mPER3). Our mPER crystal structures will guide the design of His and Cys mutants to evaluate heme binding in vivo and in vitro.

All our mPER crystal structures contain a conserved functional NES (28) within the αE helix (Fig. S4). Interestingly, mutation of the C-terminal Met469 of mPER2 to lysine significantly affects the nuclear export activity of the NES toward a heterologous protein, whereas mutation of the equivalent C-terminal Leu408 of mPER1 to Lys has no significant effect on the activity of the mPER1-NES (28). This difference might be related to the fact that Met469 is involved in mPER2 homodimer interactions, whereas the equivalent Leu residues of mPER1 and mPER3 are completely surface exposed due to the changed relative orientation of the monomers (Fig. S3A and ref. 27). Hence, homo- or heterodimer interactions of the mPER proteins are likely to modulate the function of this NES in the mammalian circadian clock.

Our mutant analyses revealed that the predicted helix-loop-helix region N-terminal to the PAS domains significantly stabilizes homodimers of mPER3 but not mPER2 (Fig. S5 E and F). Due to the high sequence similarity of the mPER helix-loop-helix regions (Fig. S1B), this different behavior was somewhat unexpected. We propose, that the changed monomer orientations revealed by our crystal structures (Fig. S3A) may affect the ability of the predicted helix-loop-helix segments of mPER2 and mPER3 to approach each other for homodimer interactions. Furthermore, crude modeling exercises suggest that formation of the helix-loop-helix dimer interface requires a reorientation of at least one of the two N-terminal caps. Notably, all our mPER homodimer structures are asymmetric in a sense that the N-terminal cap is less ordered and therefore more flexible in one of the two monomers. Additionally, the different structures and PAS-A interactions of the N-terminal caps of mPER1, 2, and 3 (Fig. 1 C and F, Fig. S3B) might play a role. Since the basic region that is essential for the DNA binding of basic-HLH transcription factors (32), is missing in all three mPER homologues (Fig. S1B) and mPERs have not been reported to directly bind to DNA (5, 29, 30), a regulation of mPER homo- or heterodimer formation by DNA is not to be expected. It is however possible, that (instead of homodimer interactions) the predicted helix-loop-helix region of mPER2 engages in heterodimeric interactions with other helix-loop-helix proteins. Like the ID (inhibitor of DNA binding) family of helix-loop-helix proteins, mPERs could thereby potentially compete with the formation of transcriptionally active and DNA-binding dimeric basic-HLH transcription factors (33, 34).

Based on the relatively low homodimer affinity of its PAS domains and the higher flexibility of its PAS-A domain and N-terminal cap, mPER2 seems more predisposed for heterodimeric signaling interactions than mPER1 and mPER3. Consistently, the PAS domains of mPER2, but not mPER1, have been reported to interact with the β-subunit of casein kinase 2 (35), glycogen synthase kinase 3β (36), REV-ERB and possibly other nuclear receptors (23, 35–37). Since the mutation of the mPER1 ³³⁵PCCLL coactivator motive to the mPER2 sequence (LCCLL) did not restore mPER1 binding to REV-ERB (23), additional regions or molecular features of mPER2 appear to be required. Due to the higher flexibility of the N-terminal cap and the ßD-ßE loop (Fig. S3 A and B), the LxxLL motive of mPER2 is likely to be more accessible for interactions with nuclear receptors. Furthermore, the two PAS-A coactivator motives in the mPER homodimers are inappropriately positioned to jointly bind to an active nuclear receptor homodimer (38). Hence, the lower homodimer affinity and enhanced flexibility of the mPER2 PAS domains may facilitate REV-ERB binding in a nonhomodimeric conformation. Apart form the PAS domains, the C-terminal mPER2 region, which also contains an LxxLL coactivator motive (aa 1050–1054), may contribute to nuclear receptor binding.

In our luciferase complementation assay with the mPER2[1–1127] protein, we have obtained the strongest luciferase signals when the N-terminal half of the luciferase enzyme was fused to the N-terminal end and the C-terminal half to the C-terminal end of the mPER fragments. This implies that, while full-length mPER2 homodimers are predominantly stabilized by the PAS-B/tryptophan interface (Fig. 2B), their N- and C-terminal ends are near each other in a cellular context. Notably, this configuration could bring N- and C-terminal LxxLL coactivator motives into a spatial arrangement that enables binding to REV-ERB homodimers or other dimeric nuclear receptors. Furthermore, the N-terminal mPER2[1–330] fragment has been shown to interact with the C-terminal mPER2[1056–1257] fragment and with the CK2 β-subunit under Co-IP conditions, suggesting that the N- and C-terminal mPER2 regions communicate with each other and jointly provide a platform for CK2 interaction (35).

In our FRAP assays (Fig. 3), mPER1 and mPER2 showed a slower fluorescence recovery (lower mobility) than tetrameric Venus-fused ß-galactosidase (MW = 580 kDa). Since Venus-mPER1- and Venus-mPER2 homodimers are only 325 kDa in size, the mPERs appear to be associated with other proteins in the cytoplasm. As mCRY1 binds directly to mPER2 in our luciferase complementation studies (Fig. 2A), the mPER complexes most likely contain mCRYs. Additional components could be the other mPER homologues and kinases such as CKI ε/δ, which were shown to be present in nuclear mPER1 and mPER2 complexes (17, 18). Interestingly, the mutation of Trp419 to glutamate leads to faster moving mPER2 complexes (Fig. 3). We conclude, that the PAS-B/Trp interface and hence the intact mPER2 homodimer is essential for the formation of the more slowly moving and presumably larger mPER2 complexes in mammalian cells. The fact that wild-type mPER2 complexes move more slowly and are therefore likely to be larger than wild-type mPER1 complexes, provides another example of the functional differences between these two mPER homologues and might be correlated with the more versatile molecular interactions of mPER2 (see above). Of course we cannot exclude that interactions with other cellular structures such as cytoskeletal proteins contribute to the delayed fluorescence recovery of the mPER complexes. Such contributions become more significant as the size of the complexes increases (39) and are, if at all, more likely to affect the wild-type mPER2 complexes.

mPER3 shares some properties with mPER2 (PAS domain homodimer with similar affinity, mostly stabilized by the PAS-B/Trp interface) and others with mPER1 (homodimer structure more mPER1-like, PAS-A/αC interface discernable). As mPER1, mPER3 possesses a second interface that raises its homodimer affinity to the sub-μM range, but different from mPER1, mPER3 uses the helix-loop-helix motive located N-terminal to the PAS domains and not the PAS-A/αC interface as additional stabilizing interface. These unique features of mPER3 are likely to be critical for its molecular interactions e.g. with other clock proteins (5, 7, 8) or the nuclear receptor PPAR-γ (24). They are therefore expected to affect its functions in clock- and sleep regulation and adipogenesis (25, 26) as well as other as yet not very clearly defined roles of mPER3 in peripheral tissues and within output pathways of the clock (40).

In light of the emerging picture that nontranscriptional metabolic rhythms reciprocally interact with the transcriptional oscillator (3, 4), the interactions of mPERs with proteins involved in metabolic and signaling functions as well as with heme might provide possible interconnections between the cellular metabolism and the clockwork. Our PAS domain crystal structures of the three mammalian PERIOD homologues and the quantitative analyses of their homodimer interactions in solution and inside cells provide plausible explanations for their distinct molecular interactions and will guide the design of mutations or small molecule ligands to further dissect their different functions within or outside the mammalian circadian clock.

Materials and Methods

A detailed description of the materials and methods (cloning, protein expression and purification, crystallization and X-ray structure determination, biochemical, spectroscopic and cell biological studies) is provided as supplemental information.