CaMKIIα mediates feedback effects of wheel-running activity on the circadian pacemaker in mice Authors: Tomoko Yoshikawa, Ken-ichi Honma, Isao Tokuda, Sato Honma Published: 2026 ABSTRACT CaMKIIα functional knockout mouse (K42R) showed substantial changes in the entrainment pattern of running-wheel (RW) rhythm to a short (LD6:18) but not to a long (LD18:6) photoperiod. In wild type mice (WT), entrainment to LD6:18 occurred at the light-off phase, whereas in most K42R entrainment took place at the light-on phase. These differences are well explained by lengthening of circadian period (τ) and compromising the couplings between the activity-onset and -end in K42R. Indeed, τ in K42R and WT were longer and shorter than 24 h, respectively, under the following constant darkness (DD). Without RW τ was lengthened in WT, whereas it was shortened in K42R. Similarly, activity time was lengthened in WT but shortened in K42R. Importantly, τ without RW was not different between the two genotypes, which were confirmed in naive mice. Tau and α returned to the pre-removal levels after RW re-introduction in both WT and K42R. These results indicate that the changes of behavioral rhythm in K42R are due to a lack of feedback effects of RW activity, which are mediated by CaMKIIα-dependent mechanisms. A two-oscillator model assuming involvement of CaMKIIα in the coupling between them successfully simulated the effects of RW activity in K42R.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1038/s41598-026-54563-4. Introduction Ca2+/Calmodulin-dependent protein kinase IIα (CaMKIIα) is an essential component for the intracellular Ca2+ functions in the brain, including the circadian behavioral rhythms. The kinase-dead mutation in mice (K42R) lengthened the circadian period (τ) of running-wheel (RW) activity rhythm, increased the activity time (α), i.e., the interval between the activity-onset and -end, and showed recurrent α decompression and α compression during free-running under constant darkness (DD)1,2. These behavioral phenotypes in K42R were interpreted as results of compromised circadian oscillation underlying the activity-onset and weakened the coupling from the activity-onset to -end2. On the other hand, entrainment to LD12:12 seemed to be unaffected in K42R1,2. Entrainment to a long and short photoperiod is explained by the non-parametric model, in which τ, α and the shape of phase response curve (PRC) for a light pulse play critical roles3. In nocturnal rodents, the circadian behavioral rhythm was compressed under a long photoperiod such as LD18:6, whereas the behavioral rhythm became decompressed under a short photoperiod such as LD6:184. In addition, entrainment of behavioral rhythm in mice whose τ was shorter than 24 h was achieved at the light-off phase where the entraining light signal produces phase-delay shifts of the circadian rhythm4, while entrainment in albino rats whose τ was longer than 24 h occurred at the light-on phase where the light signal phase-advances the circadian rhythm5. If the non-parametric entrainment model is applicable to the entrainment in K42R whose τ is longer than 24 h1,2, the entrainment to LD6:18 would be substantially different from that in WT. It is well documented that RW activity feedbacks onto the circadian pacemaker and shortens τ in nocturnal rodents6–8. RW activity given at the subjective afternoon is also known to phase-advance the behavioral rhythm9,10 and facilitates the re-entrainment to phase-advanced LD cycle11,12. However, the mechanism of feedback effect of RW activity is not well understood. In the previous studies for K42R, the behavioral rhythms were monitored by a RW actogram. It is therefore important to clarify whether the observed effects of functional loss of CaMKIIα are mediated through the feedback effects of RW activity. In the present study, we explored the issues above described in K42R, examining their entrainment patterns to LD18:6 and LD6:18, which were followed by exposure to DD. Behavioral rhythms were monitored both using RW and passive infrared sensor (IR), which are a stress and non-stress type actograms, respectively13. K42R showed altered entrainment to a short photoperiod compared to wild-type mice (WT), differences of which attributable to their lengthened circadian period. During free-running in DD, RW was removed from a cage and behavior was monitored only by IR. Surprisingly, the effect of RW removal on τ was completely opposite in K42R to that in WT. Tau was shortened in WT, whereas it was lengthened in K42R. The features of behavioral rhythm in K42R were successfully simulated by the model of coupled evening (E) and morning (M) oscillators, based on the previous suggestions14. Results Entrainment to LD18:6 and LD6:18 Entrainment of circadian behavioral rhythms to LD12:12 or to LD18:6 was not much different in pattern between the two genotypes (Fig. 1, Suppl Fig. 1 and 2). However, the entrainment to LD6:18 was substantially different between them. The behavioral rhythm in WT showed either gradual (Type a) or immediate (Type b) phase-advance shifts and kept the stable phase-relation with the light-off phase in such a way that the activity-onset was located immediately after that phase (Fig. 1A). The ratio of Type a and Type b mice was 5:3. The mean α at the end of LD6:18 was 12.52 ± 0.71 h (mean ± SD, n = 8) and the mean τ in the following DD was 23.69 ± 0.18 h (n = 8).Fig. 1Double-plotted behavioral rhythms in LD18:6 and LD6:18. (A) Two representative sets of RW (left) and IR (right) activity rhythms in WT. (B) Three representative RW activity rhythms in K42R. The mice were entrained to LD12:12, LD18:6, and LD6:18, and released into DD. Behavioral activity (counts) was expressed in 6-min bins. Vertical lines indicate the transition of light condition (light-off or light-on). Double-plotted behavioral rhythms in LD18:6 and LD6:18. (A) Two representative sets of RW (left) and IR (right) activity rhythms in WT. (B) Three representative RW activity rhythms in K42R. The mice were entrained to LD12:12, LD18:6, and LD6:18, and released into DD. Behavioral activity (counts) was expressed in 6-min bins. Vertical lines indicate the transition of light condition (light-off or light-on). In contrast, the behavioral rhythms in the majority of K42R failed to establish stable entrainment to LD6:18 and showed four different types (Fig. 1B, Suppl Fig. 2). In Type c, the activity-onset gradually phase-advanced to approach the light-off phase, whereas the activity-end did not phase-shift for the first several weeks and then started to phase-delay when the activity-onset turned to phase-delay shifts. Three K42R showed this type of entrainment and their mean α at the end of LD6:18 was 15.81 ± 1.24 h (n = 3) and the τ in the following DD was 24.16 ± 0.11 h. In Type d, the activity-onset showed gradual phase-advance shifts for the first several days but soon phase-delayed shifts. By contrast, the activity-end made phase-delay shifts from immediately after released into LD6:18. Four K42R showed this type of entrainment and their mean α at the end of LD6:18 was 14.28 ± 0.91 h (n = 4) and the mean τ in the following DD was 24.18 ± 0.06 h (n = 4). In Type e, the activity-onset immediately phase-advanced and kept a stable phase-relation with the light-off phase throughout LD6:18, whereas the activity-end showed phase-delay shifts from the very beginning of LD6:18 and kept a stable phase-relation with the light-on phase. Two K42R showed this type of entrainment and α at the end of LD6:18 was 17.93 h and 13.79 h, and τ in the following DD was 24.01 and 24.08, respectively. Two K42R showed entrainment in similar way to WT (Type f). Their α were 13.79 h and 15.45 h and τ were 23.93 h and 23.78 h, respectively. These values were similar to or slightly exceeded the highest levels in WT. The mean α of all K42R was 15.05 ± 1.46 h (n = 11), which was significantly longer than that of WT (12.52 ± 0.71, n = 8, p = 0.0003). The mean τ in the following DD was also significantly longer in K42R (24.10 ± 0.15 h, n = 11) compared to WT (23.69 ± 0.18 h, n = 8, p = 0.0006). Effects of RW removal on the circadian period of IR rhythm The behavioral rhythms of three K42R were not included in this analysis, because they showed the recurrent α decompression and compression2 after RW removal. The effects of RW removal on the circadian behavioral rhythm in K42R were completely opposite to those in WT (Fig. 2, Suppl Fig. 3 and 4). Since the τ of activity-onset and -end were not significantly different in the same individuals for the first two sessions, τ of the activity-onset was used in the following analyses. Tau in WT was significantly lengthened from 23.69 ± 0.18 h to 24.02 ± 0.09 h (n = 8, p = 0.0023), whereas that in K42R was significantly shortened from 24.17 ± 0.08 h to 24.05 ± 0.08 h (n = 8, p = 0.0019). Re-introduction of RW in WT significantly shortened τ to 23.80 ± 0.10 h (n = 8, p = 0.0012), and in K42R significantly lengthened to 24.29 ± 0.16 h (n = 8, p = 0.0010; Fig. 2B).Fig. 2Effects of RW removal on activity rhythm in DD. (A) Representative double plotted IR and RW activity rhythms are illustrated in WT (left) and K42R (right) in DD. (B) The free-running periods in three sessions with RW, without RW, with RW again are demonstrated individually (closed grey circle) and the mean (open circle) with SD by a vertical line in WT (left) and K42R (right) mice. (C) Activity time (α) in the three sessions. ** P < 0.01, * P < 0.05; repeated-measures two-way ANOVA with post-hoc Tukey–Kramer test. Effects of RW removal on activity rhythm in DD. (A) Representative double plotted IR and RW activity rhythms are illustrated in WT (left) and K42R (right) in DD. (B) The free-running periods in three sessions with RW, without RW, with RW again are demonstrated individually (closed grey circle) and the mean (open circle) with SD by a vertical line in WT (left) and K42R (right) mice. (C) Activity time (α) in the three sessions. ** P < 0.01, * P < 0.05; repeated-measures two-way ANOVA with post-hoc Tukey–Kramer test. RW removal changed α in an opposite direction for WT and K42R (Fig. 2C). In WT, RW removal significantly lengthened α from 14.22 ± 1.75 h to 15.55 ± 1.43 h (n = 8, p = 0.0119) and RW re-introduction significantly shortened α to 14.28 ± 1.54 h (n = 8, p = 0.0182). In K42R, RW removal significantly shortened α from 15.55 ± 0.87 h to 14.17 ± 0.99 h (n = 8, p = 0.0125) and RW re-introduction significantly lengthened α to 15.55 ± 0.79 h (n = 8, p = 0.0041). α without RW was significantly shorter in K42R than in WT (p = 0.0416). IR activity rhythm in the naive K42R in DD IR activity rhythms were monitored in naive WT and K42R in DD (Fig. 3A). The mean τ in WT was 23.81 ± 0.08 (n = 9) and that in K42R was 23.76 ± 0.12 h (n = 8), which were not significantly different between the genotypes (p = 0.3293) (Fig. 3B). The mean α was also not significantly different between the two genotypes (WT: 13.20 ± 0.94 h, K42R: 12.67 ± 0.74 h; p = 0.2200) (Fig. 3C).Fig. 3Free-running periods (τ) and activity times (α) in naive mice. (A) Representative double plotted IR activity rhythms are illustrated in WT (left) and K42R (right). Vertical lines indicate the transition of light condition (light-off or light-on). (B), (C) Free-running periods (B) and activity times (C) of IR activity rhythms in WT and K42R mice in DD. The animals had no access to RW and were never used for any experiments. τ and α are demonstrated individually by small open circles (WT) and small closed circles (K42R). The mean values are indicated by a large open (WT) or closed circle (K42R) with SD by a vertical line. Free-running periods (τ) and activity times (α) in naive mice. (A) Representative double plotted IR activity rhythms are illustrated in WT (left) and K42R (right). Vertical lines indicate the transition of light condition (light-off or light-on). (B), (C) Free-running periods (B) and activity times (C) of IR activity rhythms in WT and K42R mice in DD. The animals had no access to RW and were never used for any experiments. τ and α are demonstrated individually by small open circles (WT) and small closed circles (K42R). The mean values are indicated by a large open (WT) or closed circle (K42R) with SD by a vertical line. Simulation of behavioral rhythms in K42R mice by a two-oscillator model Entrainment to a short and long photoperiod The present two-oscillator mode well simulated circadian behavioral rhythms in WT and K42R (Fig. 4). Under LD18:6, the activity-onset and -end immediately shifted in both mice, with the activity-onset at the transition from the light phase to the dark (LD transition) and the activity-end at the transition from the dark phase to light (DL transition) as exactly the same manner as in the real experiments (Fig. 1). In the following LD6:18, the activity-onset in WT (Fig. 4B) re-entrained to LD with transition of a few days and with phase angle difference of 1 ~ 2 h (Type b), whereas the activity-end did not shift for a first few days and then showed phase-delay shifts for the several days until reaching a stable phase position. The process of re-entrainment differed for the two behavioral phases and resulted in lengthening of activity time as in a real experiment (Type b). A simulated rhythm in K42R (K42R(1)) which was assumed to possess a weak CME coupling showed the similar re-entrainment process in the activity-onset to the simulated rhythm in WT and a real experiment (Type b). By contrast, the re-entrainment of activity-end lacked marked phase-advance shifts following the initial phase-delay shifts. This feature was detected neither in the simulated nor in the real experiment in WT but similar to the real K42R in Type c. The simulated rhythm in K42R, however, lacked phase-delay shifts of both behavioral phases, which was detected in the real experiment ca. 20 cycles after the transfer to LD 6:18. In the other K42R simulation (K42R(2)) with a strong CME coupling, the activity-onset phase-advanced and immediately fixed at a phase-position with a larger ψ than that in K42R(1). The activity-end showed an initial and relatively larger phase-delay shifts for a few cycles and then the rate of phase-delay shifts became small. As a result, the activity time was substantially extended. This simulation was similar to the real experiment in Type d but lacked the marked phase-delay shifts of activity-onset.Fig. 4Mathematical model simulations. (A) In the model, the circadian system is composed of the E and M oscillators. The RW activity is assumed to feedback onto the circadian system (i.e., intrinsic periods, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{E}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{M}$$\end{document}, and coupling strength, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{EM}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{ME}$$\end{document}). The effects on \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{E}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{EM}$$\end{document} are mediated by a CaMKIIα dependent mechanism, while those on \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{M}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{ME}$$\end{document} are mediated by a CaMKIIα independent mechanism. (B-D) The oscillators in the model were entrained to LD18:6 for the first 10 days, LD6:18 for the next 10 days, and then exposed to constant dark conditions (C, D). Between 30 and 40th days, no RW was applied. Peak phases of the E and M oscillators are drawn in red and blue, respectively. (E) Free-running periods with RW, without RW, with RW again. (F) Activity times with RW, without RW, with RW again. ** P < 0.01; repeated-measures two-way ANOVA with post-hoc Tukey–Kramer test. Mathematical model simulations. (A) In the model, the circadian system is composed of the E and M oscillators. The RW activity is assumed to feedback onto the circadian system (i.e., intrinsic periods, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{E}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{M}$$\end{document}, and coupling strength, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{EM}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{ME}$$\end{document}). The effects on \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{E}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{EM}$$\end{document} are mediated by a CaMKIIα dependent mechanism, while those on \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{M}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{ME}$$\end{document} are mediated by a CaMKIIα independent mechanism. (B-D) The oscillators in the model were entrained to LD18:6 for the first 10 days, LD6:18 for the next 10 days, and then exposed to constant dark conditions (C, D). Between 30 and 40th days, no RW was applied. Peak phases of the E and M oscillators are drawn in red and blue, respectively. (E) Free-running periods with RW, without RW, with RW again. (F) Activity times with RW, without RW, with RW again. ** P < 0.01; repeated-measures two-way ANOVA with post-hoc Tukey–Kramer test. Free-running and effects of RW removal Under following DD, the both activity-onset and -end of the simulated rhythm in WT showed circadian periods shorter than 24 h in steady state, whereas those in the K42R demonstrated circadian periods longer than 24 h, as observed in the real experiments (Fig. 2). RW removal and reintroduction changed the circadian period of simulated rhythms in both genotypes in an opposite direction as illustrated in Fig. 4E. This was exactly the same as observed in the real experiment (Fig. 2B). The effects of RW removal on the activity time was also opposite for the simulated rhythms in WT and K42R (Fig. 4F), reflecting the results of real experiments (Fig. 2C). Importantly, in the simulation, the tau without RW was not different between WT and K42R (Fig. 4), as observed in naive mice (Fig. 3). Discussion K42R showed unique circadian behavioral rhythms under LD6:18 and in DD when monitored by RW actogram. These features included: (i) lengthened system circadian period (τsys), (ii) extended α, (iii) recurrent α decompression and compression with attenuated RW activity, (iv) unstable entrainment to LD 6:18, and (v) opposite responses of τ and α to RW removal and re-introduction compared to WT. All these features are well explained in terms of the E and M oscillator model underlying the activity-onset and -end. RW activity strengthens the E and M oscillation and the bidirectional couplings between them. The effects of RW activity on the E oscillator and the coupling from the E oscillator to M are mediated by a CaMKIIα involved mechanism. Entrainment to LD18:6 and LD6:18 Entrainment to photoperiods is explained by the non-parametric entrainment model, in which the E and M oscillators are involved15. The E oscillator is entrained by light signals hitting the subjective evening, which phase-delays the oscillator. By contrast, the M oscillator is entrained by light signals in the subjective morning, which phase-advances the oscillator16. The light responsiveness of the E and M oscillators is different in terms of PRC for a light pulse5,16. The entrainment pattern of behavioral rhythm to LD6:18 depends on τsys, which is integrated by the E and M oscillators14. Entrainment to LD18:6 was established by immediate phase-delay shifts of the activity-onset and phase-advance shifts of the activity-end, resulting in extreme α compression (Fig. 1). The phase-shifts by two pulses per day guarantee stable entrainment of the circadian system3. However, taking small amplitudes of the delay and advance portions of PRC into account, the mouse PRC for lights does not completely explain such immediate phase-shifts. Therefore, the entrainment to LD18:6 could be shaped by light masking and/or by light-induced modification of the couplings between the E and M oscillators17. In contrast, entrainment to LD6:18 is not established by the two-pulse phase-shifts due to a long dark period. The entrainment in WT was achieved by phase-delay shifts of the E oscillator, which drives the activity-onset near the light-off phase. Whereas, because an entraining light signal did not hit the light sensitive portion of PRC for the M oscillator, the phase of the activity-end was determined by the coupling strength between the E and M oscillator. The entrainment of WT was exactly the case of one-pulse phase-shifts. By contrast, entrainment of K42R was highly variable and categorized into four types, which depended on τsys and the coupling strength. Alpha under LD6:18 was significantly longer in K42R than in WT, indicating that the coupling between the E and M oscillators became weaker in K42R. The mice showing Type c and Type d entrainments had longer τsys and a wider α. The initial phase-advance shifts of activity-onset was a process of α decompression4. The phase-delaying shift of activity-end in Type d was also due to α decompression. Upon reaching a steady state phase-relation with the activity-end, the activity-onset started to phase-delay, showing the bouncing-like phenomenon similar to that observed under skeleton photoperiods3. Our interpretation of this unique phenomenon is the free-running of behavioral rhythm with τ longer than 24 h when α decompression ceased and the steady state phase relation was established between the E and M oscillators. In the behavioral rhythm of Type c, CEM is much stronger than CME and vice versa in the rhythm of Type d. The mutual coupling in Type e was very weak and the E and M oscillators were entrained to the light-off and light-on phases separately. In some animals, the phase-advance shifts looked instantaneous, and it remains possible that this shift was due to the removal of negative masking by light. RW activity is known to accelerate phase-advance shifts of the activity-onset11. A pulse of RW activity has been known to phase-advance the behavioral rhythms in hamsters when applied at the subjective evening, but not at the subjective morning9. The mechanism of RW induced phase-advance shifts is not well understood. The transient period for re-entrainment could be a temporal dissociation between the E oscillator and the activity-onset18. RW activity may facilitate the synchrony between them12. In this effect of RW, CaMKIIα may not be involved, because the immediate phase-shift was detected in both WT and K42R. Feedback effect of RW activity on the circadian system The most interesting and important findings of the present study is the opposite effects of RW activity on τsys and α in K42R. Removal of RW lengthened τ and α in WT, but shortened them in K42R. In other words, RW activity shortened τ and α in WT, but it lengthened them in K42R. In the E and M oscillator model, τsys is a function of τE and τM as well as CEM and CME. Shortening of τsys is achieved either by acceleration of the E oscillator to further shorten τE or strengthen the CEM to increase the impact of the E oscillator on τsys. Alternatively, the τsys could be shortened by inhibition of the M oscillator to decrease τM or by weakening CME to make the CEM relatively strong. However, the former possibility is more likely, because an interesting covariation was reported between τsys and distribution of RW activity. For example, castration lengthened τ and produced an ill-defined onset of activity in male mice, whereas testosterone replacement shortened τ and produced sharp onsets of activity19. In K42R during free-running in DD, when τsys increased, RW activity became more prominent at the activity-end2. These findings suggest that RW activity at the activity-onset affects the E oscillatior to shorten τsys. The effects of RW activity are referred as a non-photic time cue and they affect the circadian system in an opposite way from a light pulse7,8. The non-photic entrainment is also interpreted in terms of effects of RW activity on the E oscillator, most likely on the neural couplings of constitutional cellular oscillators for the E oscillator. There is a line of evidence that the regional pacemakers in the SCN are composed of multiple cellular oscillators4,20,21. Role of CaMKIIα in the feedback effect of RW activity Previously, CaMKII was suggested to mediate the light-induced phase-delay shifts of circadian behavioral rhythm. Chemical inhibitors for CaMKII attenuated the light-induced phase-delay shifts22–24. Light signals stimulated the SCN neurons through glutamate secretion, which activated NMDA receptors and subsequently CaMKII through Ca ion influx24. CaMKII then phosphorylated CREB to induce the clock gene Pers transcription, resulting in phase-shifts of the circadian pacemaker. However, this scenario explains neither the entrainment to a long and short photoperiods, nor the feedback effect of RW activity in mice lacking functional CaMKIIα. Behavioral rhythms in K42R entrained to LD12:12 and LD18:6 similarly to WT, in which phase-delay as well as phase-advance shifts were required in response to light signals3. The present results indicated that the feedback effect of RW activity on the E oscillator and on CEM was mediated by a CaMKIIα dependent mechanism. Without functional CaMKIIα, RW activity failed to shorten τ and to maintain the stable phase-relation between the activity onset and end1,2(Fig. 2 in this paper). Importantly, a loss of functional CaMKIIα did not affect these circadian parameters when behavioral rhythm was monitored by IR. This result was confirmed in the naive WT and K42R which had never been used for other experiments. Neither τ nor α were significantly different between the two genotypes. Therefore, CaMKIIα is necessary for the RW activity to activate the E oscillator or strengthen CEM. However, a lack of CaMKIIα lengthened τ beyond the level of IR activity rhythm in WT, indicating that RW activity enhances the M oscillator and/or CME as well, which are not affected by CaMKIIα disfunction. RW activity is thought to stimulate the M oscillator and its coupling to the E by different mechanism from those for the E oscillator and it’s coupling to the M. This interpretation is based on phenotypic observations, and further investigation of the underlying mechanisms is necessary for confirming its validity. In the previous studies1,2, τ of the PER2::LUC rhythm in the cultured SCN slices was significantly longer in K42R than in WT. These findings are not consistent with the present behavioral results. The reason for this discrepancy is unknown. The circadian PER2::LUC in the SCN culture may not directly reflect the IR activity rhythm21. Regional circadian pacemakers have been reported in the ventral and dorsal regions of the SCN25–27 and in the rostral and caudal regions of the SCN4,21,28. The E and M oscillators was previously suggested to locate in either of the bilateral SCN (31) but functional redundancy of the bilateral SCN was shown in behavior29 and entrainment30. There is a line of evidence that the vasopressin (AVP) neural networks in the SCN play a critical role in the coupling between the regional oscillators31–33. Furthermore, there is evidence that the AVP-mediated coupling is bidirectional but not symmetrical34. CaMKII was suggested to mediate AVP production in the paraventricular nucleus in the hypothalamus35. It is possible to assume that RW activity stimulates the AVP networks involved in the E oscillator formation and signal transductions from the E to M oscillator. The current findings are all from male mice, because of sex dependent differences in the activity. Circadian activity rhythms of female mice are known to show cyclic changes that parallel their estrous cycles36. Further study including female mice may be necessary to generalize the findings. A mathematical model of the circadian rhythm in K42R mice The most important finding of the present study is the feedback effect of RW activity on the circadian pacemaker, which is mediated by the CaMKIIα dependent mechanism. In our model, RW activity is assumed to affect the functions of both the E and M oscillator in WT with further assumption that the CaMKIIα dependent mechanism affects the E oscillator but not the M (Fig. 4). These assumptions are based on the findings that the coupling between the two oscillators is bidirectional and affected differently by experimental manipulation31,37. The model successfully simulates the opposite effect of RW removal on the as well as the activity time in K42R. Namely, RW activity strengthens functions of the E oscillator such as τE and CEM more than those of the M oscillation in WT, whereas it enhances exclusively the M oscillator functions in K42R which lose the CaMKIIα dependent mechanism mediating the effect of RW activity to the E oscillator. The model also simulated the entrainment to a long and short photoperiod. Four types of entrainment patterns in LD6:18 were detected in K42R; two of them (Type c, d), the popular types (7/11), were satisfactory simulated by the model but the other one (Type e), a less popular type (2/11), was not simulated. We need further investigation to simulate the type e entrainment by, e.g., improving the shape of PRC which was in the present model set to be symmetric for the advance and delay portions and the same for WT and K42R. Most rodent species, the advance and delay portions are asymmetric and related to the system τ5,15. Conclusions K42R show a variety of phenotypes in circadian behavioral rhythms when measured by RW actogram. Their rhythms are explained by a lack of RW activity which mediated feedback effect on the E oscillator and the coupling from the E to M oscillator. Materials and methods Animals Animal experiment was conducted in accordance with the Guidelines for the Care and Use of Laboratory Animals of Hokkaido University. The present experiments were approved by the Animal Research Committee of Hokkaido University (permission number: 16–0073). CaMKIIα knock-in mouse of C57BL/6J background (K42R) was created by introducing point mutation that replaced the 42nd lysine to arginine38. Heterozygote K42R pairs were used for breeding to obtain homozygous K42R and littermate WT. All WT used in this study were the littermate WT, and they were derived from the same litters as K42R. Animals were bred and raised in our animal facility under LD12:12 with light-on at 06:00 h and -off at 18:00 h. The light intensity during the light phase was about 100 lx (fluorescent light). Room temperature was 22 ± 2 °C, and humidity was 60 ± 10%. Chow (Oriental Yeast Co., Tokyo, Japan) and tap water were available at all times. Weaning was done at the postnatal day 21. Male mice of homozygous K42R and littermate WT type mice were used for the present experiments. Measurement of behavior activity and data analyses At the age of 2 months old, male mice were transferred to individual cages equipped with a running-wheel (RW, diameter 10 cm) in a light-tight box where light intensity was approximately 300 lx in the light phase and less than 0.01 lx in the dark. The light intensity was measured by a luminometer (TOPCON/IM-5, Tokyo), which measures the light intensity from 0.001 – 1,999,000 lx. Animal cares (replenishment of food pellets to a feeding box, of drinking water and cage exchange) in constant darkness (DD) were done with an aid of infrared LED light (850 nm). The animals were exposed to this infrared light every 7–10 days, which lasted less than a minute. We monitored two different behaviors of mice, general movement and RW activity as described elsewhere13. Briefly, general movements were measured with a passive infrared (IR) sensor (KPS-3227, ELEKIT) attached to the ceiling of recording box. The distance between the sensor and the bottom of the cage was 29 cm. The sensors detect the change in thermal radiation (infrared wave) emitted from a mouse, which is due to the locomotion of the animal. Temporal resolution of the detection was 4 s. The IR sensors do not respond to small movements such as rearing or grooming. Simultaneously, the numbers of wheel revolutions (full turn) were counted. IR and RW activities were recorded every min with a PC system (Stanford Chronobiology kit, Stanford Software Systems, CA, USA) and analyzed with Clock Lab (Actimetrics, IL, USA). Activity data were analyzed in 6 min bins. Tau of behavioral rhythm was measured separately for the activity onset and end under steady state free-running. Activity onset and end of each cycle were visually defined in the double-plotted actogram, then, a regression line fitted to consecutive activity onsets or ends were drawn by Clock Lab software for at least 20 cycles. The line drawn to activity onset was used to calculate τ. The interval between the activity-onset and -end was defined as α, a length of activity band. One researcher performed the measurements, and another researcher independently replicated the analyses. Although there were slight differences in the absolute values (0.02 ± 0.11 h for τ and 0.48 ± 1.42 h for α, n = 48) , both researchers confirmed significant difference in τ and α between with and without RW and between WT and K42R. Experimental procedures Experiment 1. Entrainment to LD18:6 and LD6:18 Male mice of 2 months old were used (11 K42R and 8 littermate WT). The behavioral rhythms were monitored by RW and IR for 3 weeks in LD12:12. Then the animals were exposed to LD18:6 by delaying the light-off phase for 3 h and advancing the following light-on phase for 3 h. Therefore, the midpoint of dark or light phase was not changed in terms of local time. Thirty or 37 days later, they were exposed to LD6:18 by advancing the light-off phase for 6 h and delaying the following light-on phase for 6 h, so that the midpoint of dark or light phase was not changed. Thirty or 37 days later, they were released into DD for 6 weeks by extending the last dark period of LD6:18. Experiment 2. Effect of removal of RW on τ Following the exposure to LD6:18, the mice were released into DD for 14 weeks and the effect of RW removal on τ was examined. RW was removed for 4 weeks from a cage at 6 weeks after the release into DD. During this period, behavioral rhythm was measured by IR. Subsequently, RW was re-introduced in a cage for another 4 weeks. Tau during these three sessions (with RW, without RW and with RW again) were calculated for the activity-onset and -end separately. The measurement was done for the last 20 cycles of each session. Experiment 3. τ in RW naive K42R and WT in DD Eight K42R and 9 WT were raised in LD12:12. They had no access to RW and never used for any kind of experiment. Behavioral rhythms were monitored by IR. They were released into DD and τ was determined by Chi-square periodogram39, using data of 40 days in DD under steady state free-running. Experiment 4. Mathematical simulation Mathematical simulation was done for the experiment 1 and 2, basing on the model of two mutually coupled phase oscillators14. The oscillator driving the activity-onset was referred to as the evening (E) oscillator and that driving the activity-end as the morning (M) oscillator, according to Pittendrigh’18 (Pittendrigh, 1981). We assume intrinsic period of the E oscillator (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{E}$$\end{document}) is shorter than 24 h and that of the M (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\tau}_{M}$$\end{document}), longer than 24 h as previously proposed14. The coupling between the two oscillators is bidirectional and asymmetric. The coupling from the E oscillator to the M (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{EM}$$\end{document}) is assumed stronger than the opposite direction (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${C}_{ME}$$\end{document})40. Statistical analyses Statistical analyses were performed using Excel Tokei 2012 (SSRI, Japan). The difference between the genotypes was evaluated by t-test, and the differences among different sessions were by repeated-measures two-way ANOVA with post-hoc Tukey–Kramer test. Supplementary Information Supplementary Information. Supplementary Information.