FRQ is a large, dynamic component of the clock oscillator

(A) The TTFL in Neurospora crassa. In the core circadian oscillator, FRQ associates with FRH and CK1 to form a repressor complex (the FFC) This complex inhibits the positive-acting transcription-factor known as the white-collar complex WCC, composed of White Collar-1 (WC-1) and White Collar-2 (WC-2). The WCC activates frq expression by binding to the c-box region in the frq promoter. Stoichiometries of the components are not represented in the schematic. (B) AlphaFold derived model of LFRQ color coded by pLDDT score that highlights the confidence in each predicted region. (C) Functional domains of FRQ showing the start position for the LFRQ isoform (AUGLFRQ.) and the four native cysteine residues. Characterized domains include: CC – coiled coil domain, FCD – FRQ:CK1 interacting domain, PEST – proline, glutamic acid, serine, and threonine-rich domain and FFD – FRQ:FRH interacting domain.

Purification of FRQ in non-phosphorylated and phosphorylated forms reveals IDP behavior

(A) SDS-PAGE gel showing non-phosphorylated (np-FRQ) and phosphorylated FRQ (p-FRQ) alongside a molecular weight ladder. Both samples represent the purified full-length long isoform (L-FRQ, residues 1-989) in non-denaturing buffer. (B) Size exclusion chromatography–multiangle light scattering (SEC-MALS) of phosphorylated FRQ (p-FRQ) and non-phosphorylated (np-FRQ). The molecular weight (MW) of p-FRQ (∼143 ± 1 kDa) is somewhat larger than that of an unphosphorylated monomer (∼120 kDa), but substantially smaller than that of a dimer, whereas the high concentration MW value of np-FRQ (∼261 ± 1 kDa) is slightly larger than that expected for a dimer (∼240 kDa). Note that the p-FRQ MW is affected by >80 phospho sites. The shoulder to the right of the main peak had low light-scattering and a large MW error. The colored lines represent the MW of each sample. (C) Dimensionless Kratky plot of SAXS data for p-FRQ and np-FRQ, in both cases the peak positions of the curves are shifted from that expected for a globular protein (√3,1.104), but neither plateau at 2, which would be characteristic of a flexible, denatured polypeptide (Durand, et al., 2010). See Table S1 for additional SAXS parameters. Total protein concentration was between 50-75 μM for the MALS and SAXS experiments.

ESR spectroscopy studies of spin-labeled FRQ.

(A) Spin-labeling strategies for FRQ. MTSL was used for internal labeling of cysteine residues and AEP1 and SortaseA were used to label the N and C terminus, respectively. (B) AlphaFold generated model of FRQ with the positions of native spin-labeled cysteine residues highlighted in dark blue. (C) X-band (9.8 GHz) room temperature cw-ESR measurements for various spin-labeled FRQ variants, each containing a single Cys labeling site, as designated. Data represented as first-derivative spectra with the horizontal axes depicting the static field strength (typically 3330 to 3400 G), and the vertical axis depicting the change in magnetic susceptibility with respect to the static field, in arbitrary units. (D) Time-domain data of DEER measurements for double-cysteine labeled FRQ variants. The baseline-corrected time domain dipolar evolution trace fits are shown; details are given in Fig. S5-6. The 490-658 separation is noticeably closer and more ordered than the others. The total protein concentration was between 50-75 μM for all experiments.

Properties of the FFC.

(A) SEC of the FFC; left inset: SDS-PAGE gel inset of the components; right inset: 32P-autoradiogram of FRQ phosphorylation by CK1 and CK1 autophosphorylation at 10 and 20 μM FRQ all in the presence of FRH. Complete autoradiograms are shown in Fig. S7. (B) Overlay of dimensionless Kratky plots of SAXS data from the FFC formed with either np- or p-FRQ (protein concentrations were between 50-75 μM). The SAXS-derived MW of np-FRQ FFC = 285 kDa, which compares to a predicted value of 273 kDa; MW of p-FFC = 340 kDa; Table S1).

PDS distance distributions of spin-labeled FRQ and the FFC.

(A) DEER-derived distance distributions of the FFC wherein the spins were directed by ADP linkage to target the FRH and CK1 ATP-binding sites (blue) or in the form of MTSL labeling of cysteine residues within the FCD2 and FFD domains of FRQ (red). Time-domain data and distance distributions with error estimates are shown in Fig. S5-S10. (B-G) DEER-derived distance distributions of cysteine residue pairs labeled with MTSL within p-FRQ, either alone (blue) or within the FFC (red). (H) Percentage change in the average separation of FRQ when it binds FRH and CK1. Errors are derived from the uncertainties in the distance distributions shown in Fig. S9-10. In these experiments, only FRQ was labeled with MTSL at the positions noted. Note: Total protein concentrations were between 50-75 μM for all experiments.

FRQ undergoes LLPS in agreement with predictions from its sequence properties and those of its functional analogs.

(A) LLPS propensity predictions using “Pi-Pi” and IUPred2A disorder predictions are shown for FRQ (purple), dPER1 (teal) and hPER1 (orange) as compared to values for proteins of similar length from their respective proteomes. The mean and standard deviations are represented as horizontal and vertical black bars, respectively. Note: **** denotes a p-value <0.0001 obtained from a Mann-Whitney U-test (B) DIC microscopy images of various concentrations of p-FRQ (in 150 mM NaCl, 25 mM Tris pH 8.0) and np-FRQ (in 500 mM NaCl, 25 mM Tris pH 8.0) at 25°C; images shown at 100x magnification. Scale bar = 2μm (C) Temperature vs concentration phase diagram derived from the results of the turbidity assays shown in Fig. S17. Both p-FRQ and np-FRQ undergo an LCST phase transition on or above the line (a second order polynomial fit to the data points). The error bars of each point (about the size of the points) reflect the 95% confidence interval of the mean. The phase boundary for p-FRQ represents the transition apparent in the temperature scans of Fig. S17. Some p-FRQ phase separation already occurs at low temperature under these buffer conditions and thus the region below the line does not represent fully soluble p-FRQ. (D) Live cell fluorescent imaging of FRQmNeonGreen in nuclei of Neurospora hyphae. Upper left panel shows the FRQ channel, represented as a multicolor heat map of fluorescence intensity; upper right panel shows a surface plot derived from the raw FRQ image to emphasize regions of concentration; lower left shows nucleoporin SON-1mApple, which localizes to the cytoplasmic face of the nuclear envelope; bottom right shows the FRQ:SON-1 merge image. The images were acquired on a Zeiss 880 laser scanning confocal microscope and were smoothed by bicubic interpolation during 10-fold enlargement from 42 × 42 pixels to 420 × 420 pixels. (E) FRQ[mNeonGreen] nuclear foci shown within an N. crassa syncytium mycelium (outlined by white lines), cropped from movie S1. Scale bar = 2μm.

Conditions favoring LLPS alter the structural and enzymatic properties of the FFC.

(A) Differential Interference Contrast (DIC) and fluorescence microscopy images at 25 °C of phase separated np-FRQ or p-FRQ (5 μM each) droplets in 500 mM KCl, 20mM Na2HPO4, pH 7.4 with either equimolar Cy5-labeled FRH, CK1 or CheY (control) visualized using Cy5 fluorescence. Scale bar = 2μm . (B) (Top) Autoradiography of FRQ (20 μM) (left) phosphorylated by CK1 in the presence of FRH at increasing temperatures. (Bottom) Schematic showing the nature and results of the autoradiography assay depicted above. The FRQ phosphorylation under LLPS conditions was reduced relative to non-LLPS conditions. The complete autoradiograms are shown in Fig. S13. Quantification of FRQ phosphorylation by CK1 at RT under LLPS and non-LLPS conditions with phosphorylation levels under non-LLPS conditions normalized to 1. Note: *** denotes a p-value << 0.05 obtained from a student’s t-test. (C) DIC microscopy images of 15 μM phase separated np-FRQ under the same conditions (i.e. buffer and temperature) as the phosphorylation assay shown in (B). Scale bar = 2μm . (D) X-band (9.8GHz) RT cw-ESR spectra of FRQ labeled with MTSL at the 490 site in solubilizing buffer (black) and under conditions that promote LLPS (cyan). Data represented as first-derivative spectra with the horizontal axes depicting the static field strength (typically 3330 to 3400 G), and the vertical axis depicting the change in magnetic susceptibility with respect to the static field, in arbitrary units.

SAXS parameters for p-FRQ, np-FRQ and their reconstitution into the FFC

FRQ Cysteine residue locations and context

FRQ regions that bind clock proteins

ο values from the cw-ESR central peak for FRQ variants of Fig. 3C

R-square values for the quality of the two component gaussian fits to the distance distributions.

Calculated values from the UV-Vis turbidity assays of np-FRQ shown in Figure 6C.

Calculated values from the UV-Vis turbidity assays of p-FRQ shown in Figure 6C.

(A) AlphaFold model of LFRQ colored by pLDDT score. (B) Predicted error alignment plot that shows the uncertainty in the relative positions of two residues in the structure. (Image source: https://www.uniprot.org/uniprotkb/P19970/entry#structure)

Domain map of FRQ with phospho-sites (as identified by mass-spec) highlighted. The sites in red were also identified in (Baker, et al., 2009; Tang, et al., 2009).

Pair-wise distance distributions calculated from SAXS data for p-FRQ (red), np-FRQ (dark blue), p-FFC (magenta), and np-FFC (light blue).

Enzymatic labeling of FRQ at its termini. X-band (9.8 GHz) RT cw-ESR measurements for the spin-labeled peptide by itself and when it has been grafted onto either the N or C-terminus of FRQ.

Time domain traces before (Raw) and after (Data) background correction and trace fits (Fit) from DEER experiments of different double-cysteine FRQ variants alone or within the FFC (i.e., when FRQ was bound to FRH and CK1a). The spins were present on the residue number of FRQ that is highlighted for example ‘370-490’ refers to FRQ spin-labeled at the 370C and 490C site and the DEER experiment probed the dipolar coupling of the spins at these sites. The last graphs show the time domain data for the control single cysteine DEER experiments with p-FRQ 490C and p-FRQ 746C.

Baseline-corrected time domain fit traces from Q-band DEER experiments of different double-cysteine FRQ mutants alone (blue) and within the FFC (orange). Original data are shown in Supplementary Figures S5 and S9.

Full autoradiograph of np-FRQ phosphorylation by CK1 and γ-32P-ATP from Figure 4A inset. Four technical replicates are shown for each np-FRQ concentration.

Background-corrected time domain trace and error analyses from DEER experiments targeting the ATP-binding pockets of FRH and CK1. The DEER experiment probed the dipolar coupling of the spins in the ADP-β-S-SL molecule that was present in ATP-binding pocket of FRH or CK1.

Error analysis of the distance distributions produced from DEER experiments of different double-cysteine FRQ variants. The spins were present on the amino acid number of FRQ that is highlighted for example ‘370-490’ refers to FRQ spin-labeled at the 370C and 490C site and the DEER experiment probed the dipolar coupling of the spins at these sites.

Error analysis of the distance distributions produced from DEER experiments of different double-cysteine FRQ variants within the FFC (i.e., when FRQ was bound to FRH and CK1). The spins were present on the amino acid number of FRQ that is highlighted for example ‘370-490’ refers to FRQ spin-labeled at the 370C and 490C site and the DEER experiment probed the dipolar coupling of the spins at these sites.

(Left)The two component Gaussian fitting of the distance distribution (P[r]) obtained from SVD for FFC 370-490 (i.e., spins were placed on resides 370 and 490 on FRQ, and this protein was bound to CK1 and FRH). No constraints were placed on the width, position, or amplitude of either component and the best fit values were obtained by minimizing the difference between the fits and the SVD distance distributions. Similar fits were carried out for all the different variants. The quality of the fit is reported by the R-square values and ranged between 0.94-0.99 for all the samples tested. They are reported in Table S5. (Right) The proportion of the closer distance peak in the distance distributions of FRQ (blue) or FRQ+FRH+CK1(FFC) (red) samples. The error bars reflect the uncertainty in estimating the population of each peak and were derived from uncertainties in the distance distributions shown in SI Appendix, Fig. S9 and S10.

FRQ, but not FRH or CK1, is highly disordered and predicted to undergo LLPS. Shown are the disorder and LLPS propensity scores of the members for the FFC complex, FRQ, FRH and CK1. Columns from left to right are disorder and phase separation predictions for FRQ, FRH and CK1, respectively. Disorder graphs (red) were computed using the IUPred2A program (Mészáros, et al., 2018) where values above a 0.5 threshold (dashed line) are considered disordered. LLPS propensity predictions were computed using “Pi-Pi” (blue) (Vernon, et al., 2018) and catGRANULE (light blue) (Bolognesi, et al., 2016). Dashed lines represent neutral propensity. Shown on the top right corner of each graph are the percent disorder or LLPS propensity scores of each protein.

FRQ and its functional homologs are predicted to undergo LLPS. Shown are the disorder and LLPS propensity scores for FRQ, dPER1 and hPER1. Data for the disorder graphs (red) were computed using the IUPred2A program where values above a 0.5 threshold (dashed line) are considered disordered. Pi-pi predictions (blue) were computed using (Vernon, et al., 2018), while catGRANULE predictions (light blue) were computed using catGRANULE server (Bolognesi, et al., 2016). Shown on the top-right corner of each graph are the percent disorder, PScore or catGRANULE score for each protein. Dashed lines represent neutral propensity. FRQ, dPER1 and hPER1 are all predicted to be highly disordered (86.77%, 58.84% and 63.07%, respectively) and undergo LLPS.

FRQ and its homologs, dPER1 and hPER1, have similarly disparate physicochemical properties compared to their proteomes. (A) Shown are the distribution of fraction of charged residues (FCR), net charge per residue (NCPR), and hydropathy along the sequence of FRQ, dPER1 and hPER1. (B) Sequence characteristics of FRQ and its homologs, dPER1 and hPER1. Various protein sequence parameters are calculated using localCIDER (Holehouse, et al., 2017) and displayed including fraction of charged residues (FCR), the degree of charge mixing (Kappa), net charge per residue (NCPR), and hydropathy (the hydrophobic and hydrophilic character of protein chain as defined by (Kyte, et al., 1982)) for the UniProtRef50 clusters of FRQ (purple), dPER1(blue) and hPER1 (orange). For comparison are shown protein sequences of similar length from the SWISS-Prot reviewed proteomes of Neurospora crassa (light purple), Drosophila melanogaster (light blue) and Homo sapiens (light orange). Hydropathy is consistently low for the clock proteins and FRQ orthologs have a high degree of charge separation, i.e., tracks of negative and positive residues (high kappa). High kappa values are common for proteins that phase separate (Somjee, et al., 2020). (C) Classification of the charge properties of the Ref50 clusters of FRQ and its functional homologs, dPER1 and hPER1. Das-Pappu phase plots showing the fraction of negatively vs positively charge residues for the UniProtRef50 clusters of FRQ (purple), Neurospora crassa proteome control (light purple), dPER1 (blue), Drosophila melanogaster proteome control (light blue), hPER1 (orange) and Homo sapiens proteome control (light orange). Shown are the fraction of negative residues versus fraction of positive residues for each cluster of homologs and their controls (similar size proteins from their respective proteomes). Despite the lack of amino acid conservation between the functional homologs of FRQ, dPER1 and hPER1, the ratio of fraction negative to fraction positive residues are similar among the clock proteins. dPER and hPER1 have a similarly low value of both fraction positive and fraction negative compared to their respective proteomes, whereas the fraction positive and fraction negative residues are both higher in FRQ. Note: Ref50 is a clustering method utilized by UniProt and is defined in (Steinegger, et al., 2018). While this work was in preparation, conceptually similar analyses were carried out by Jankowski et al., 2022 (Jankowski, et al., 2022).

FRQ and its functional homologs dPER1 and hPER1 have similar amino acid composition compared to their respective proteomes. FRQ and its functional homologs dPER1 and hPER1 have similar amino acid compositions, which differs substantially from that of their respective proteomes. The amino acid composition for FRQ, dPER1 and hPER1 and their respective homolog clusters and proteomes are shown. Generally, G, S, T and P residues are enriched whereas hydrophobic (A, I, L, V) residues are relatively depleted for FRQ, dPER1, hPER1 and their homologs compared to their proteomes. Unlike dPER1 and hPER1 and their homologs, FRQ and its homologs are also highly enriched in R and D.

Phosphorylation modulates the physicochemical properties of FRQ. Shown are the distribution of fraction of charged residues (FCR), net charge per residue (NCPR), and hydropathy along the sequence of np and p-FRQ (top panel). In the bottom panel, shown are the disorder and LLPS propensity scores for np and p-FRQ. Data for the disorder graphs (red) were computed using the IUPred2A program (Mészáros, et al., 2018) where values above a 0.5 threshold (dashed line) are considered disordered. Pi-pi predictions (blue) were computed using, while catGRANULE predictions (light blue) were computed using catGRANULE server (Bolognesi, et al., 2016). Shown on the top-right corner of each graph are the percent disorder, PScore or catGRANULE score for each protein. Dashed lines represent neutral propensity. Note: Glu(E) residues were employed as phospho-mimetics in all analysis.

Exploring the phase behavior of FRQ. (A) UV-Vis absorption spectrum of p-FRQ (in 150 mM NaCl, 25 mM Tris pH 8.0) and np-FRQ (in 500 mM NaCl, 25 mM Tris pH 8.0) as a function of temperature at increasing concentrations. Note: the shading around each curve represents the standard deviation from the mean absorbance from 2 technical replicates. 15 μM p-FRQ had the greatest standard deviation as seen by the largest shaded area. (B) UV-Vis absorption spectrum of p-FRQ and np-FRQ in a non LLPS promoting buffer as a function of temperature. Note: p-FRQ appears to undergo a small amount of LLPS even in non-LLPS buffer as seen by its relatively high absorbance relative to np-FRQ.(C) (Left) UV-Vis absorption spectrum of p-FRQ (in 150 mM NaCl, 25 mM Tris pH 8.0) and (Right) np-FRQ (in 500 mM NaCl, 25 mM Tris pH 8.0) as a function of temperature to check for reversibility of LLPS transition. (D) DIC micrograph of FRQ showing droplets fusing and docking as indicated within the yellow circles. (E) UV-Vis absorption spectrum of p-FRQ (Left) and np-FRQ (Right) as a function of temperature in either low salt (150 mM) or high salt (500 mM).

Full radiograph of np-FRQ phosphorylation by CK1 and γ-32P-ATP from Figure 7.

Schematic diagram summarizing a potential mechanism for LLPS mediated circadian clock regulation. By reducing CK1 activity, especially at higher temperatures, LLPS provides a potential temperature compensation mechanism for the clock.

Visualization of FRQ [mNeonGreen] in living Neurospora, displaying heterogeneous, punctate intranuclear localization. Images of a single focal plane were acquired in a time series with 300 ms between exposures.