Preservation of circadian rhythms by the protein folding chaperone, BiP

Dysregulation of collagen synthesis is associated with disease progression in cancer and fibrosis. Collagen synthesis is coordinated with the circadian clock, which in cancer cells is, curiously, deregulated by endoplasmic reticulum (ER) stress. We hypothesized interplay between circadian rhythm, collagen synthesis, and ER stress in normal cells. Here we show that fibroblasts with ER stress lack circadian rhythms in gene expression upon clock‐synchronizing time cues. Overexpression of binding immunoglobulin protein (BiP) or treatment with chemical chaperones strengthens the oscillation amplitude of circadian rhythms. The significance of these findings was explored in tendon, where we showed that BiP expression is ramped preemptively prior to a surge in collagen synthesis at night, thereby preventing protein misfolding and ER stress. In turn, this forestalls activation of the unfolded protein response in order for circadian rhythms to be maintained. Thus, targeting ER stress could be used to modulate circadian rhythm and restore collagen homeostasis in disease.—Pickard, A., Chang, J., Alachkar, N., Calverley, B., Garva, R., Arvan, P., Meng, Q.‐J., Kadler, K. E. Preservation of circadian rhythms by the protein folding chaperone, BiP. FASEB J. 33, 000–000 (2019). www.fasebj.org

Circadian clocks are cell-autonomous timekeeping mechanisms that occur in most tissues to optimize cellular activities in anticipation of varying demands on the cell during 24 h (1, 2). One such function is control of translation and protein synthesis (3). In a recent study, Yeung et al. (4) showed that the circadian clock regulates translocon-dependent endoplasmic reticulum (ER) protein synthesis and membrane trafficking at each node in the secretory pathway of tendon fibroblasts in vivo. The consequence is a daily surge of type I collagen (collagen-I) production, which occurs at night in mice. Collagen-I is the major secreted protein of fibroblasts and, in tendon, accounts for 70% of the mass of the tissue, where it occurs as extracellular fibrils (5) that can extend the length of the tissue (6). Collagens are large trimeric molecules that undergo extensive post-translational modification (7,8) and folding in the ER (their site of synthesis) in order to generate a thermally stable triple helix capable of assembling into fibrils. The extensive nature of post-translational modification of collagen, coupled with a surge in synthesis at night, suggested that the cell might have mechanisms to protect against the accumulation of unfolded collagen [e.g., with the unfolded protein response (UPR)] and subsequent ER stress at times of high collagen synthesis. Excessive collagen synthesis occurs in fibrosis and solid tumors, where the rate of synthesis exceeds turnover. The result is accumulation of a dense, stiff extracellular matrix that can be a positive fibro-proliferative feedback signal to accelerate cell division and the synthesis of additional matrix (9).
Collagen-I molecules are composed of 2 a1 chains and 1 a2 chain encoded by the genes collagen-I a1 chain (COL1A1) and collagen-Ia2 chain (COL1A2), respectively. These chains come together to form a triple helix, which is assisted by chaperone proteins. Both generic protein chaperones, such as binding immunoglobulin protein [BiP, also known as heat shock protein (Hsp)A5] or 78-kDa glucose-regulated protein (Grp78) and Grp94, and the collagen-specific chaperone, Hsp47, aid the formation and stabilization of the triple helix (10). Chaperones also have signaling roles associated with their ability to bind newly synthesized proteins. In particular, BiP is known to govern the activation of the UPR (11). When unfolded proteins accumulate in the ER, BiP is released from 3 key binding partners on the ER membrane: protein kinase R-like ER kinase (PERK), inositol-requiring enzyme 1 (IRE1), and activating transcription factor (ATF)6 each activate mechanistically distinct pathways. Combined, the response acts to reduce ER burden through mRNA decay and reducing translation but also drives expression of protein chaperones. However, if the stress is not resolved, these same pathways can also promote apoptosis (12). It has been established that the response to ER stress is regulated in a circadian manner in cancer cells (13) and that activation of UPR in cancer cells induces a 10 h shift in circadian oscillation (14); however, it remains unknown how protein flux through the ER, and in turn ER stress, regulates circadian rhythm in nontumorigenic models. Given that collagen secretion is ramped up from low levels during the day to high levels in the night in mice (4), we hypothesized that the expression of these chaperones may be governed by circadian rhythm. In the experiments described below, we explored the connection between ER stress and circadian rhythm in fibroblasts actively secreting collagen.

Luminometry
LumiCycle apparatus (Actimetrics, Wilmette, IL, USA) was used for real-time quantitative bioluminescence recording. Tail tendons, 10 mg, were placed in 30-mm cell culture inserts (0.4 mm pore size; MilliporeSigma, Burlington, MA, USA) inside 35-mm dishes in recording medium [DMEM without phenol-red (D2902; MilliporeSigma), supplemented with 4 g/L glucose, 5% fetal calf serum, HEPES, sodium bicarbonate, and 0.1 mM luciferin]. iTTFs were seeded 24 h, as previously described, prior to drug treatments. Dexamethasone (100 nM) was added for 30 min to synchronize the circadian rhythms before plating in recording medium. Baseline subtraction was carried out using a 24-h moving mean. Amplitude was calculated as peak-trough difference in bioluminescence of the first and second peak, as indicated, using baseline-subtracted data.

Dose response curves
The appropriate dosages of thapsigargin and tunicamycin for in vitro treatments were determined in triplicate using the Ala-marBlue assay (Thermo Fisher Scientific, Waltham, MA, USA) in 96-well format (20).

Real-time PCR
For quantitative PCR analysis, primers designed using Assay Design Center (Roche, Basel, Switzerland), the sequences used are indicated in Supplemental Table S1. Reactions were run using Power-Up Sybr Master Mix (Thermo Fisher Scientific) with cDNA generated using a TaqMan Reverse Transcription Kit (Thermo Fisher Scientific). RNA (0.5 mg) extracted using Trizol (Thermo Fisher Scientific) was used to generate cDNA, which was then diluted 20-fold for PCR. PCR reactions were run on the CFX96 Real-Time System (Bio-Rad, Hercules, CA, USA). Target transcript expression was normalized to the geometric mean of ribosomal protein lateral stalk subunit P0 (RPLP0), Gapdh, and actin b.

Immunofluorescence and in-cell Western blot
For immunofluorescence analysis cells were fixed with 4% paraformaldehyde and permeabilized with 0.5% Triton-X-PBS. Collagen-I was detected using a rabbit pAb (1:200; OARA02579; Gentaur), and ER was identified using a mouse mAb against protein disulfide isomerase (PDI) (1:100; ab190883; Abcam, Cambridge, MA, USA). Images were collected on a Leica TCS SP5 Acousto-Optical Beam Splitter inverted confocal using a 360/0.50 Plan Fluotar objective and 33 confocal zoom (Leica Microsystems, Buffalo Grove, IL, USA). The confocal settings were as follows: pinhole 1 Airy unit, scan speed 1000 Hz unidirectional, format 1024 3 1024. Images were collected using photomultiplier tube detectors with the following detection mirror settings: FITC 494-530 nm; cyanine 5 (Cy5) 640-690 nm using the 488 nm (20%), 594 nm (100%), and 633 nm (100%) laser lines, respectively. To eliminate cross-talk between channels, the images were collected sequentially. When acquiring 3dimensional optical stacks, the confocal software was used to determine the optimal number of z sections. Only the maximum intensity projections of these 3-dimensional stacks are shown in the results. For imaging nonhelical collagen, cells were permeabilized as above and stained with anticollagen-I (1:200, OARA02579; Gentaur) and 20 mM 5-carboxyfluorescein-conjugated collagen hybridizing peptide (3Helix, Salt Lake City, UT, USA). Cells were stained for 2 h at 4°C. Images were collected on a Zeiss Examiner A1 upright microscope using a 363/1.4 N-Achroplan objective (Carl Zeiss, Oberkochen, Germany) and captured using a Coolsnap ES camera (Photometrics, Huntington Beach, CA, USA) through Metavue v.7.8.0.0 software (Molecular Devices, Sunnyvale, CA, USA). Specific band pass filter sets for DAPI, FITC, and cyanine 3 (Cy3) were used to prevent bleed-through from 1 channel to the next. For in-cell Western blots, iTTFs were grown in 96-well plates in the presence of 200 mM ascorbic acid; 24 h after plating, cells were treated with tunicamycin and thapsigargin as indicated before removing treatment and adding full medium supplemented with ascorbic acid and dexamethasone. After 72 h, cells were fixed with 4% paraformaldehyde. To assess extracellular collagen, cells were directly stained with an anti-collagen antibody (1:1000; OARA02579; Gentaur) overnight, which was detected using donkey anti-rabbit 800 antibody (1:10,000, 5151s; Cell Signaling Technology, Danvers, MA, USA); each well was counterstained with the cell-permeating dye deep red anthraquinone 5 (DRAQ5). The plate was scanned using the Odyssey CLx (Li-Cor Biosciences, Lincoln, NE, USA). The collagen signal was quantified following normalization to deep red anthraquinone 5. To assess intracellular collagen, the plates were permeabilized as for immunofluorescence studies before being stained and quantified in the same manner as extracellular collagen.

Statistics
For assessment of data fit to circadian rhythms, microarray data were analyzed using the MetaCycle package for R (https://CRAN. R-project.org/package=MetaCycle), which combines 3 approaches (JTK_Cycle, Lomb-Scargle, and Arser) to decide on rhythmicity. Protein data were assessed using CircWaveBatch v.3.3 (https:// www.euclock.org/results/item/circ-wave-batch.html). Where indicated, paired Student's t tests were performed using triplicate analysis from at least 2 independent experiments. A value of P , 0.05 was deemed significant.

BiP is circadian clock rhythmic
BiP is known to interact with collagen in the secretory pathway (21) and is proposed to aid the folding of collagen. As a first experiment, we showed that levels of BiP are rhythmic with a 24-h period (Metacycle Benjamini-Hochberg q-value 0.003) and its peak of expression precedes the peak of expression of pro-collagen-I (the precursor of collagen) protein (Fig. 1A). The circadian nature of BiP protein expression was also observed in mouse embryonic fibroblasts (MEFs), isolated tail tendon fibroblasts, and tail tendon tissue with peak and minimal expression ( Fig. 1B and Supplemental Fig. S1). The fluctuations in BiP levels during the day support recent observations demonstrating that BiP is a short-lived protein that is turned over in 2-3 h (22). Examination of chromatin immunoprecipitation sequencing data sets (23) showed that the Hspa5/BiP promoter is differentially bound by the core clock component cryptochrome protein 2 (Cry2) during the course of 24 h, with RNA polymerase II being recruited at times corresponding to elevated BiP transcription (Supplemental Fig. S1) This result was the first indication that the circadian clock had a role in preemptively averting ER stress by up-regulating BiP in anticipation of a surge in collagen synthesis.

The presence of misfolded protein ablates circadian rhythm
To learn more about the effects of ER stress on the circadian rhythm, we treated the tail tendons of ex vivo Per2::luc clock reporter mice (2) with thapsigargin. Thapsigargin inhibits the sarcoplasmic or ER Ca-ATPase family, which leads to a reduction of ER calcium levels and so induces protein misfolding and ER stress (24). When cells were subjected to dexamethasone to synchronize the circadian clock, the amplitude of rhythm was markedly reduced after thapsigargin treatment (Fig. 1C). Similarly, when iTTFs from Per2::luc mice were pretreated for 1 h with thapsigargin or tunicamycin (an inhibitor of N-linked glycosylation, which normally aids protein folding) at doses that do not impact on cell growth (Fig. 1D), the amplitude of Per2::luc rhythm was dampened (Fig. 1E).
This connection between ER stress and circadian rhythm was further exemplified in cells that were in the process of resolving ER stress prior to dexamethasone treatment. We used a 5-h treatment with thapsigargin or tunicamycin to accumulate misfolded proteins in the ER and induce all arms of the UPR (Supplemental Fig. S2) prior to synchronization of the circadian rhythm. This treatment regime eliminated the synchronization effect of dexamethasone and forskolin ( To establish if misfolded proteins instigate the dampening of circadian rhythms, we used a mutated thyroglobulin that misfolds in the ER, thereby causing ER stress (25). When this mutated thyroglobulin was stably expressed in Per2::luc cells, there was dramatic dampening of circadian rhythm (Fig. 1G, H) and induction of ER stress, particularly increasing the splicing of X-box binding protein 1 (XBP1) (Fig. 1I). Thus, the fact that circadian rhythm is lost in fibroblasts expressing misfolded thyroglobulin shows that the loss of rhythm is the direct consequence of accumulated misfolded protein.

Suppression of the secretory pathway blocks circadian rhythm by induction of ER stress
The effect of ER stress induction on collagen secretion was examined by Western blot analysis of the conditioned medium of thapsigargin-or tunicamycin-treated cells using an anti-collagen-I antibody. Over 48 h, both thapsigargin and tunicamycin abolished secretion of collagen-I ( Fig. 2A). To examine if blockade of the secretory pathway had a similar effect on the secretion of collagen, cells were treated with brefeldin A and monensin (Fig. 2B, C). These treatments suppressed collagen secretion and induced ER stress, as shown by induction of CCAAT-enhancer binding protein-homologous protein (CHOP) transcription after 5 h treatment (Fig. 2D). These effects were comparable to those of thapsigargin and tunicamycin, and likewise, both brefeldin A and monensin dampened circadian fluctuations in Per2::luc cells (Fig. 2E).
Having demonstrated that short-term treatment with thapsigargin or tunicamycin dampens circadian rhythm, we sought to utilize this to assess the importance of circadian rhythm on the secretion and assembly of collagen fibers. First, we showed that after  treatment with 100 nM dexamethasone, there is enhanced fibril assembly in iTTFs (Supplemental Fig.  S4); these short-term effects of dexamethasone have previously been demonstrated in refs. 26 and 27 in fibroblasts. We used a modified in-cell Western blot approach to demonstrate the effects of thapsigargin and tunicamycin on collagen secretion (Fig. 2F, G). As expected, continuous treatment of fibroblasts with ER stress inducers reduced the assembly of extracellular collagen (Fig. 2F), which was accompanied by accumulation of intracellular collagen (Fig. 2G), in line with reduced secretion of collagen into the medium (Fig. 2A). Having validated the in-cell Western blot, we then induced ER stress using a 5-h pulse of thapsigargin or tunicamycin prior to dexamethasone treatment. The results showed that disruption of rhythm reduced the accumulation of extracellular collagen after 72 h, implicating circadian rhythms in coordinating collagen secretion and assembly. Importantly, the intracellular levels of collagen were unaffected by this treatment regime, indicating that collagen had not been held in the ER. A 1-h treatment did not affect either extracellular or intracellular collagen. We have evaluated collagen fiber assembly in fibroblasts isolated from the circadian locomotor output cycles Kaput D19 (ClockD19) N-ethyl-N-nitrosoure (ENU) mutation mouse, a mutation which is a deletion in exon 19 of the Clock gene resulting in diminished circadian activity (28). These fibroblasts lack a sustained Per2::luc rhythm in response to dexamethasone treatment (Fig. 2H). In cultures of ClockD19 fibroblasts, there were increased numbers of collagen fibers compared with wild-type fibroblasts (Fig. 2I, J) and reduced ER stress-related transcripts (Supplemental Fig. S5), which supports recent observations that showed increased collagen deposition in the tendons of ClockD19 mice (4). The deposition of collagen fibers by fibroblasts fluctuates during the day (4). These timedependent changes in collagen are not observed in cultures of ClockD19 fibroblasts (Fig. 2I), suggesting that circadian rhythm is required for normal collagen fiber homeostasis. Together these results suggest that ER stress can regulate circadian rhythm and that circadian rhythm affects collagen fiber homeostasis, but that collagen secretion per se does not require an intact circadian rhythm.

BiP retains collagen in the ER but maintains circadian rhythm
Given the circadian rhythmicity of BiP protein levels, which peaks ahead of collagen-I (see Fig. 1A), we examined how BiP overexpression affects collagen-I secretion and the circadian rhythm. BiP was overexpressed in iTTFs using retroviral transduction (Fig.  3A), which was confirmed by real-time PCR and Western blotting (Fig. 3B, C). BiP overexpression led to inhibition of collagen fiber assembly at 2 time points following synchronization (Fig. 3D, E). Both dexamethasone and serum shock induced a significant increase in collagen fibers in synchronized cultures (Supplemental Fig. S4), implying that synchronization can aid the assembly of collagen fibers. Forskolin had the opposite effect (Supplemental Fig. S4); although forskolin can synchronize the circadian clock, it also increases cAMP levels, which has previously been demonstrated to inhibit collagen trafficking (29) and TGF-b induced collagen production (30). Based on this, we have observed that addition of the cellpermeable cAMP analog can suppress collagen synthesis in the presence of dexamethasone (Supplemental Fig. S4). In control cells, fiber assembly was observed at times after BiP levels had peaked (see Fig.  1B). Overexpression of BiP did not alter the transcription of either Col1a1 or Col1a2 (Fig. 3F), suggesting that the reduced assembly of collagen fibers is a result of altered collagen transport through the secretory pathway. Immunofluorescence detection of intracellular collagen in BiP-overexpressing cells showed that the majority of the collagen colocalizes with the ER marker PDI, whereas in control cells, regions of the ER are clear from collagen. Thus, collagen is retained in the ER when BiP is overexpressed (Fig.  3G). The use of collagen-hybridizing peptide (31) indicated that the collagen held in the ER of BiPoverexpressing cells is largely unfolded (Supplemental Fig. S6). Therefore, we examined the impact of BiP overexpression on circadian rhythm. To our surprise, the cells have a stronger inherent rhythm (Fig. 4A) and increased amplitude of Per2::luc signals following dexamethasone treatment (Fig. 4B, C). In a recent genome-wide small interfering RNA screen (32), knockdown of BiP was shown to suppress circadian rhythm (Supplemental Fig. S7). BiP is well known to suppress the activation of all 3 arms of the UPR; therefore, this result implies that the activation of the UPR arms is responsible for the dampening of circadian rhythms in cells undergoing ER stress. Real-time PCR analysis showed reduced levels of the ATF6 and IRE1-XBP1 arms of the UPR in BiP-overexpressing cells; however, there is enhanced activation of the ATF4 arm, as shown by elevated CHOP expression (Fig. 4D).
Protein folding in the ER provides a checkpoint at which circadian rhythm can be rapidly controlled The influence of unfolded proteins on circadian clock regulation was explored using clinically approved chemical chaperones, which aid the folding of proteins within the ER (33). 4-Phenylbutyric acid (4PBA) and ursodeoxycholic acid (UDCA) induced a visible rhythm in fibroblasts but also greatly increased the amplitude of Per2::luc signals following dexamethasone treatment (Fig. 5A-F). These results resemble the effects observed with BiP overexpression and suggest that misfolded proteins are resident in the ER of cultured cells and, through promoting their folding, can rescue dampened circadian rhythms. The chaperone inhibitor 2-phenylethynesulfonamide had the opposite effect, reducing Per2::luc amplitudes, which was most likely due to the observed induction of the UPR (Supplemental Fig. S8). Assessment of the ER stress pathways following 4PBA treatment showed that there is reduced expression of both ATF6 and IRE1-XBP1 arms of the UPR, again mirroring the effects of BiP overexpression, and although ATF4 expression was enhanced, there was no elevated expression of CHOP (Fig. 5G). Treatment of tendon fibroblasts with 4PBA also led to enhanced secretion of collagen fibers (Fig. 5H, I) but without altering transcription of collagen-I (Fig. 5J). Taken together, these findings imply that protein folding in the ER is a rate-limiting step in collagen biosynthesis and provides a checkpoint for control of the circadian clock.

DISCUSSION
The intrinsic circadian timing mechanism allows a cell to anticipate and adapt to daily rhythmic changes in physiologic demand. In this study we have shown that BiP is rhythmic in tendon and cultured fibroblasts and that the peak of BiP expression occurs just ahead of the peak of pro-collagen-I expression. We further showed that ER stress-inducing agents or expression of misfolded ER proteins weakens circadian rhythm and decreases collagen secretion. The use of chemical chaperones to aid protein folding shows that the resolution of ER stress can also enhance circadian rhythm and collagen deposition.
Yeung et al. (34) showed that tendon tissue (with a composition of 80-90% collagen) has an autonomous circadian clock and identified 745 rhythmic transcripts that had a 24-h oscillation. However, transcripts encoding collagens were not rhythmic. In a subsequent study, Yeung et al. (4) demonstrated that the protein levels of collagen-I oscillate with a 24-h period, suggesting that the circadian rhythmicity of collagen-I is generated posttranscriptionally. The ramp in procollagen concentration in the ER at night might be expected to lead to ER stress if the levels of nascent unfolded protein are allowed to escalate unchecked, because the levels of BiP are also circadian-regulated and peak a few hours ahead of the peak of procollagen. This suggests that the circadian clock prepares the cell for the surge in nascent procollagen chains being synthesized. We propose that after the surge of procollagen has passed, BiP levels reduce to avoid swamping the ER stress sensors and thereby reestablish homeostasis to the UPR sensing machinery. We have demonstrated here that induction of ER stress can dampen the circadian rhythm in normal fibroblasts, suggesting that this surge of BiP may influence how cells perceive external circadian cues. The fluctuations in BiP levels appear to play a protective role in maintaining circadian rhythms, as observed with BiP overexpression. This is presumably because of the saturation of all 3 sensors of the UPR. Thus, we propose that the circadian clock has a role in preemptively preventing ER stress in normal cells, by up-regulating BiP in anticipation of a surge in protein translation. However, overexpression of BiP led to retention of procollagen in the ER, which is similar to what has been observed for other proteins (35,36), notably that the retained procollagen was in a nonhelical conformation concomitant with reduced assembly of collagen fibers. A  circadian control of BiP levels suggests that a balance is struck to have sufficient BiP to allow saturation of the UPR sensors and a robust circadian rhythm, but not too much to inhibit folding of procollagen. We have shown that BiP overexpression and chemical chaperone treatment lead to suppression of Xbp1 splicing and ATF6 transcription. This suggests that these arms of the UPR are integrated into a feedback control of the molecular clock during ER stress. In these experiments, the ATF4 arm is largely unaffected; ATF4-knockout MEFs have previously been shown (37) to have reduced amplitude of Per2::luc rhythms, suggesting this pathway is essential for the robust rhythms, which may explain why these pathways remain active while the other arms of the UPR are suppressed. Of note, we have shown that ER stress or misfolded proteins can dampen clock activation in normal fibroblasts, whereas in a recent study in cancer cells undergoing ER stress a 10 h shift in circadian rhythm was observed (14). This suggests that cancer cells may negate the suppressive role of the UPR on the clock in order to survive. It may be that in a normal cell undergoing ER stress, the uncoupling of circadian clock control might facilitate the decision to resolve stress or undergo apoptosis. It has been proposed that cancer cell survival is enhanced through suppression of Clock and the brain and muscle ARNT-like 1 protein (Bmal1) (14), and both ER stress and UPR activation are well documented in many cancers, implicating the accumulation of misfolded proteins in these cells (38). Similarly, ER stress is active in patients with idiopathic pulmonary fibrosis (39), liver fibrosis (40), and kidney fibrosis (41,42). Given that activation of ER stress would normally lead to dampening of circadian rhythms and collagen synthesis, this suggests that cells within diseased tissues have circumvented this physiologic control in order to drive fibrosis. Establishing how this is achieved in diseased tissues will be the focus of future studies.