ARTICLE

DOI: 10.1038/s41467-018-03898-2 OPEN A proteomics landscape of circadian clock in mouse liver

Yunzhi Wang1, Lei Song2, Mingwei Liu3, Rui Ge1, Quan Zhou3, Wanlin Liu3, Ruiyang Li1, Jingbo Qie1, Bei Zhen3, Yi Wang3,4, Fuchu He1,2,3, Jun Qin1,3,4 & Chen Ding1

As a circadian organ, liver executes diverse functions in different phase of the circadian clock. This process is believed to be driven by a transcription program. Here, we present a tran-

1234567890():,; scription factor (TF) DNA-binding activity-centered multi-dimensional proteomics landscape of the mouse liver, which includes DNA-binding profiles of different TFs, phosphorylation, and ubiquitylation patterns, the nuclear sub-proteome, the whole proteome as well as the tran- scriptome, to portray the hierarchical circadian clock network of this tissue. The TF DNA- binding activity indicates diurnal oscillation in four major pathways, namely the immune response, glucose metabolism, fatty acid metabolism, and the cell cycle. We also isolate the mouse liver Kupffer cells and measure their proteomes during the circadian cycle to reveal a cell-type resolved circadian clock. These comprehensive data sets provide a rich data resource for the understanding of mouse hepatic physiology around the circadian clock.

1 State Key Laboratory of Genetic Engineering, Human Phenome Institute, Institutes of Biomedical Sciences, School of Life Sciences, Zhongshan Hospital, Fudan University, Shanghai 200032, China. 2 School of Life Sciences, Tsinghua University, Beijing 100084, China. 3 State Key Laboratory of Proteomics, Beijing Proteome Research Center, National Center for Protein Sciences, Beijing 102206, China. 4 Alkek Center for Molecular Discovery, Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Department of Molecular and Cellular Biology, Baylor College of Medicine, Houston, TX 77030, USA. These authors contributed equally to this work: Yunzhi Wang, Lei Song. Correspondence and requests for materials should be addressed to F.H. (email: [email protected]) or to J.Q. (email: [email protected]) or to C.D. (email: [email protected])

NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2

he mammalian circadian clock includes a “master clock” metabolism, are found to be oscillating in this TF DBA centric within the suprachiasmatic nucleus (SCN) of hypothala- view, facilitating our further understanding of the physiological T “ ” mus and peripheral clocks within other tissues of the regulations of the circadian clock in the liver. body. The “master clock” functions as an “orchestra conductor” to direct “peripheral clocks” through yet-to-be defined pathways1, allowing animals to adapt their feeding, activity, and metabolism Results to predictable daily changes in the environment. Multi-dimensional liver proteome of the circadian clock.To Circadian clocks orchestrate physiological rhythms via the map the molecular landscape of circadian clock using mouse liver temporal regulation of gene expression to control core clock as a model system, we took advantage of several current techni- ques to measure (1) TF DBA patterns (DBAP); (2) post- genes and rhythmic output programs. A network of fi transcriptional–translational feedback loop comprised of core translational modi cations (PTM) including phosphorylation transcriptional activators (Bmal1 and Clock) and repressors (Per and ubiquitylation; (3) transcriptome, and (4) protein abundance, and Cry), to control the rhythmicity in gene expression2,3. Tp53 including the nuclear sub-proteome and the whole proteome and Myc, which are well-characterized cancer driver genes, and (Fig. 1a). We aimed to provide a multi-dimensional data set for several multi-functional nuclear receptors (NRs) including Rev- the investigation and understanding of circadian clock system – erb, Ror, and Ppar family4 6, have also been shown as important within the framework of a TF-centric paradigm. regulators of the circadian clock. These studies demonstrate the critical roles of TFs in regulating circadian rhythm. In-depth TF DBAP during circadian cycle of mouse liver.We Liver plays a fundamental role in circadian clock system. Tran- fi fi fi rst pro led the diurnal rhythm of the TF DBA (Methods section; scriptome pro ling of the liver has demonstrated the circadian Supplementary Fig. 1a). Using the TF data set reported by Ravasi variation in the expression of genes related to oxidative metabolism, et al.14, we identified 617 DNA-binding proteins (DBPs), among mitochondrial functions, and amino acid turnover7, and that tran- fi 8 which, 297 were de ned as TFs. We also detected 247 tran- scriptional regulation drives the circadian mRNA rhythms .In scription co-regulators (TCs) (Fig. 1b; Supplementary Data 1). contrast, much less is known on the protein level. With rapid 9 The same time points in the two consecutive circadian cycles development of analytical techniques ,particularlymassshowed high correlation coefficient with an average Pearson’s r of spectrometry-based proteomics, it is increasingly feasible to measure greater than 0.96 (Supplementary Fig. 1b). An average of 206 TFs proteins in order to understand the diverse biological processes. were quantified for the 16 time points, ranging from 150 TFs in Recently, Robles et al.10 and Wang et al.11 reported proteome studies 10,11 the ZT12 (Zeitgeber time) to 230 TFs in the ZT0. Oscillations of in circadian clock of the mouse liver . However, due to the TF numbers as well as numbers of DBP and TC numbers were technical limitations in proteomics techniques applied, the dynamics evident (Fig. 1c, d) during the consecutive circadian cycles. of transcription factors—the key drivers of gene regulations around the circadian clock, were still poorly understood. It is expected that a hierarchical circadian regulation network Diurnal oscillations of mouse liver transcription factors.We might exist, which may include different regulatory layers that grouped the ZT points into four time phases (TP) in 6 h intervals facilitate signal transductions around the clock. The TF DNA- based on the oscillation patterns of the TF DBA. We then selected binding activities (DBA), which play key roles in regulating tran- TFs whose DBA in any specific TP that were more than twofolds scriptome, would impact the nuclear sub-proteome and in turn, the greater than that in the rest of the TPs. These TFs may perform whole proteome; post-translational modifications, including phos- specific functions in the specific TP and were therefore named as phorylation and ubiquitylation, may also impart another layer of TP-specific TFs (Fig. 2a). GO/pathway analyses of TP-specific regulation. The complicated relationships among different layers TFs identified the following enrichments: immune response in raise many questions that remained to be answered, for instance: (1) TP1, lipid metabolism in TP2, cell cycle in TP3, and glucose how diurnal rhythmic phosphorylation of signaling transduction metabolism in TP4 (Fig. 2b), suggesting a clear division of labor regulates the rhythm of TF DBA; (2) is there correlation between for hepatic functions during the circadian cycle. nuclear TF protein expression and TF DBA; (3) how diurnal We then investigated diurnal rhythmicity of DBA of the TP- rhythmic TF DBA correlates with the diurnal rhythm of down- specific TFs. Diurnal rhythmicity was identified according to stream genes’ transcription; (4) is there correlation between diurnal JTK_CYCLE8 method (Methods), in which 159 DBPs, 80 TFs, rhythms of mRNA expression and protein expression; and (5) how and 59 TCs were found as diurnal rhythmic proteins (p < 0.1), the ubiquitylation system controls the proteome oscillation. accounting for 26%, 27%, and 24% of the total identifications in Answers to these questions will be informative in portraying a its functional group, respectively (Fig. 2c; Supplementary Data 1). panoramic view of the circadian transcription regulation that gov- Importantly, Bmal1, Clock, Cry2, and Bhlhe40 were found with erns the temporal switch of physiology in the mouse liver. significant diurnal rhythms. Interestingly, they were not found to We previously developed an approach that enables the iden- be diurnal rhythmic in protein abundance from previous tification and quantification of endogenous TFs at the proteome studies15,16 (Fig. 2d). We also found that the circadian scale. With a synthetic DNA containing a concatenated tandem- transcription activator protein complex, Clock/Bmal1, showed array of the consensus TFREs (catTFRE) as an affinity reagent, we high synchrony in DBA (Pearson’s r 0.96), while the activity of can identify almost all expressed TFs in cell lines and tissues12.In the transcription repressor Cry was negatively correlated with the this study, we employ this catTFRE approach to profile the Clock/Bmal1 complex. dynamic of the TF sub-proteome during the circadian cycle of the We identified the diurnal rhythmic TFs (p < 0.1) within TP- mouse liver13. A total of 297 TFs are quantitatively identified, and specific TFs, in which 11/48 TP-specific TFs in TP1, 8/33 TP- 80 of which show circadian oscillations (p < 0.1). In addition to specific TFs in TP2, 2/12 TP-specific TFs in TP3, and 5/41 TP- TF DBA, we also measure the nuclear sub-proteome, whole-liver specific TFs in TP4 were diurnal rhythmic, respectively. The GO/ transcriptome, whole-liver proteome, phosphorylation, and ubi- pathway enrichments of diurnal rhythmic TFs in each TP were quitylation patterns to build a TF DBA-centered multi-dimen- highly consistent with the functions enriched by the TP-specific sional proteomics landscape of circadian regulation in the mouse TFs, suggesting that the hepatic gene transcription and liver. Four major biological processes, including immune physiology functions are dominated by the diurnal rhythmic response, fatty acid metabolism, cell cycles, and glucose TFs (Fig. 2e; Supplementary Data 2).

2 NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 ARTICLE

a

0 12243648

Every 3 hours, 3 mouse livers were pooled together

Phosphorylation pattern

TiO2 Nuclear subproteome WTE TF binding activity pattern catTFRE NE Transcriptome

mRNA diGG Proteome

Ubiquitylation pattern

Physiological switch of mouse liver

b c 600 DNA-binding protein DNA-binding protein 500 TF 217 400 TC

TC TF 300 103 144 297 200 Protein number 100

0 0 6 12 18 24 3036 42 Zeitgeber time (h)

d DNA-binding protein TF TC 600 250 ZT42 200 ZT0 ZT0 ZT0 ZT42 ZT42 100 150 ZT6 ZT6 300 ZT6

ZT36 ZT36 ZT36

ZT12 ZT12 ZT12

ZT30 ZT30 ZT30

ZT18 ZT18 ZT18 ZT24 ZT24 ZT24

Fig. 1 Transcription factor centered multi-omics of circadian clock in mouse liver. a Systematic workflow of the multi-omics experiments of the circadian clock in mouse liver. b Venn diagram illustrates the number of DBPs, TFs, and TCs detected in our deep TF DNA-binding activity profiling. c, d The number of DBPs, TFs, and TCs detected at each ZT point across two consecutive cycles

As previously reported, the NRs have been known as a typical diurnal rhythms of the 7 NRs passed the JTK_CYCLE test, Mlr, diurnal rhythmic TF family on the mRNA level17,18. In our study, Vdr, and Esr1 displayed single-pulse. Compared with the we detected 36 of the 49 NRs (Supplementary Fig. 2a, previous work on mRNA level conducted by the Nuclear Supplementary Data 1), including core clock regulators such as Receptor Signaling Atlas (NURSA), we found that Pparσ,Trα, Rev-erbα and Rorα, (Supplementary Fig. 2c) and 10 NRs that Rev-erbα, and Tr2 were diurnal rhythmic on both transcription showed diurnal rhythm (Supplementary Fig. 2b). While the level and DBA level, while Rorα, Lxrβ, Rxrβ were only diurnal

NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2

a ZT0/24 b TP1 TP2 TP3 TP4

0.5 Glucose metabolism

TP4 TP1 1 Immune responseLipid metabolism Cell cycle 41 48 0.5 ZT18/42 ZT6/30 0 TP3 TP2 -score 12 33 z –0.5 ZT0/24 –1 ZT12/36

ZT12/36

c Diurnal rhythmic DBP Diurnal rhythmic TF d ZT0 ZT12 ZT24 ZT36 ZT48 ZT0 ZT12 ZT24 ZT36 ZT48 Bmal1 Clock 3 3 2 2 1 1 0 0 -score -score z z –1 –1 –2 –2 0 12 24 36 48 0 12 24 36 48 Zeitgeber time (h) Zeitgeber time (h)

Cry2 Bhlhe40 Diurnal rhythmic TC 3 3 ZT0 ZT12 ZT24 ZT36 ZT48 2 2 1 1 0 0 -score -score z z –1 –1 –2 –2 0 12 24 36 48 0 12 24 36 48 Zeitgeber time (h) Zeitgeber time (h) –2 2 e TP1 TP2 Innate immune response Steroid metabolism Defense response Neutral lipid metabolism Activation of immune response Glycerol ether metabolism Innate immune response Lipid homeostasis Immune response Triglyceride metabolism Humoral immune response Glycerolipid metaolism 01020 02 6 10 –log(p value) –log(p value) TP3 TP4

Cell cycle process Glucan metabolism M phase Cellular glucan metabolism Lipid storage Cell cycle Cellular glucan metabolism Regulation of centrosome cycle Gluconeogenesis Regulation of cell cycle process Glucose metabolism 024 0246 –log(p value) –log(p value)

Pathways enriched by diurnal rhythmic TP-specific TF Pathways enriched by TP-specific TF

Fig. 2 Diurnal oscillations of hepatic transcription factors. a Rose diagram shows the number of TP-specific TFs of each TP. b Area graph shows the normalized abundance of TP-specific TFs DNA-binding activity and the bioprocess enriched by TP-specificTFs.c Cluster heat maps of the diurnal rhythmic DNA-binding activity of DBPs, TFs, and TCs. The color bar indicates normalized z-scored iBAQ. The upper white to black bar indicates the 2 days’ cycle. Daytime is shown in white, while nighttime is shown in black. d Temporal levels of the abundance of DNA-binding activity of Bmal1, Clock, Cry2, and Bhlhe40 in two consecutive cycles. X axis represents the sampled time points, Y axis represents the relative abundance of DNA-binding activity (normalized z-scored iBAQ) of each protein. e The GO bioprocesses enriched by diurnal rhythmic TP-specific TFs (red) and by all TP-specific TFs (black) of four TPs

4 NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 ARTICLE rhythmic on the DBA level, probably because they functioned as a Diurnal rhythmic TF DBA was not originated from nuclear partner of heterodimer with regulatory genes involved in liver TF. To determine whether the diurnal rhythms of TF DBA were metabolism19,20. Vdr, which was not detected in the liver on mRNA originated from the diurnal rhythmic expression levels of the level, was detected with a single-pulse in the ZT0 and ZT24 during nuclear TFs, we measured the liver nuclear proteome. We iden- the two consecutive cycles in this study (Supplementary Fig. 2a), tified 4038 (Supplementary Fig. 3a) proteins from purified nuclei highlighting the unique value of our TFRE approach. in the two diurnal cycles including 105 TFs (Supplementary

abBmal1’s DNA-binding activity Dbp’s DNA-binding activity Pho-Bmal1’s protein expression Pho-Dbp’s protein expression 3 3 2 2 Diurnal rhythmic phosphoproteins 1 1 ZT0 ZT12 ZT24 ZT36 ZT48 0 0 -score -score z z –1 –1 Total –2 –2 1657 phosphoproteins phosphoprotein 1657 0 12 24 36 48 0 12 24 36 48 9465 phosphopeptides Zeitgeber time (h) Zeitgeber time (h) Diurnal rhythmic Rev-erbα’s DNA-binding activity phosphoproteins Clock’s DNA-binding activity 9823 phosphosites Pho-Rev-erbα’s protein expression Pho-Clock’s protein expression 89 3 3 2 2 1 1 2 −2 0 0 -score -score z z –1 –1 –2 –2 0 12 24 36 48 0 12 24 36 48 Zeitgeber time (h) Zeitgeber time (h) c Immune response Lipid metabolism Kinase Kinase Ampka1 Ck2a1 Jak2 Akt1 Pdk1 Erk2

Substrate Substrate

Ifnar2 Ifih1 Ankrd17Bcr Myd88 Pcbp2 Samhd1 Pip5k1cRbl2 Pikfyve Mtmr1 Cps1 Plcb3 p 0.00013 p 3.11e–7 p 6.9e–10 p 1.12e–14 5 Mapk1Mapk3 Map3k5 Akap13 Rps6kb1 Ctnnb1 Cps1 Lpin3 Inpp5f Stk11 Inppl1 Stat5b Sorbs1 4 Stat5b Hspb1 Cc2d1a Bcl10Smarca4 Acin1 Srpk2 4 Atp2b1 Prkaca Mlxipl Pla2g4a Ptdss1 Pcyt2 3 3 Src Pdpk1 Pla2g4aAbl1 Arhgef5 Pml Sin3a 2 Nr1d1 Hdac2 Ddx20 Nr1h3 Hnf4a Acox1 2 Shortest path Shortest path 1 1 Phase specific TF TF All TF Phase specific TF* All TF* Phase specific TF Zbtb16Pbx1 Foxp1 Nfatc1 Nfatc3 All TF Phase specific TF* All TF* TF

Rfx1 Hmgb3 Bcl6b Runx3 Irf5 Nr0b2 Nr2f2 Nr1i3 Nr1i2 Nr2f6 Pparδ

Nfκb1 Stat1Stat3 Irf1 Irf3 Ets1 Nr5a2 Nr1h4 Hnf4g Esrrg Rxrg Vdr Pparγ

Cell cycle Glucose metabolism Kinase Kinase p 0.0004 p 6.33e–8 p 4.6e–5 p 4.31e–6 4 Pkcd Ck1d 4 Akt Mtor Pkcb P70s6k 3 3 2.5 Substrate Substrate 2 Med1 Hmga2 Recql5 Atrx Cdk11b Cdkn1b 2 Baiap2l1Akt1 Akt2 Gsk3a Cad 1.5 Shortest path

Shortest path 1 Eif4ebp2Pck1 Irs1 Irs2 Prkag2 Phase specific TF Eif4ebp1 Gsk3a Atf2 Crebbp Trp53bp1 Appl1 1 All TF Phase specific TF* All TF* Phase specific TF All TF Phase specific TF* All TF* Ncor2 Scap Prkcd Mtor Pak4 Kif13a Dlg1 Apc Prkcd Akt1

Rps6Mybbp1a Per2 Adam17 Foxk1 Bad

TF TF

α DbpSmad4 Smad3 Arntl Clock Nr1d2 Ror Mef2c Srebf1

Peak time p value Day (this study) p<0.1 q<0.1 Phosphoproteins detected in Day (Robles et al.) both this study and Robles et al. Night (this study) p<0.1 * Data from Robles et al. Night (Robles et al.) q<0.1

NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2

Data 3). While 79 were detected in the TFRE (DBA profiling) Nomination of the dominant rhythmic TFs.AsTFDBAmay data set, only 10 TFs (p < 0.1) were diurnal rhythmic based on the dictate downstream gene transcription, we set out to correlated the JTK_CYCLE analysis (Supplementary Fig. 3b). The Pearson’s diurnal rhythmic mRNA oscillation with the perspective of TFs. We correlation coefficient was low (r = 0.179) between TF DBA and profiled the transcriptome during the circadian cycles and found TF protein abundance (Supplementary Fig. 3c). Consistent with that 2349 of the 11,120 identified mRNAs showed diurnal rhythm the pervious study10, Bmal1 and Clock did not demonstrate (p < 0.05), which include the core clock TFs Bmal1, Per1,andDbp diurnal rhythmicity on the protein expression level. We then (Fig. 4a, b; Supplementary Data 6). Similar to the TF DBA, the performed parallel reaction monitoring (PRM) and western rhythmic gene transcripts were enriched in the immune response, blotting on several TFs protein expression level to validate these cell cycle, lipid metabolism, and glucose metabolism. However, the findings. Protein expression levels of TFs such as Bmal1/Arntl, enrichment of these four bioprocesses were not apparent in a 4-TP pparδ, Rorα did not demonstrate oscillations throughout the rhythm, but in a 2-group rhythm (Fig. 4c; Supplementary Data 7), circadian cycle (Supplementary Fig. 3d, e). Thus, pronounced with lipid metabolic and immune processes in the daytime group, diurnal oscillation of TF DBA is unlikely originated from oscil- and cell cycle and glucose metabolic processes in the nighttime lation of TF abundance during the circadian cycle. group. This pattern of behavior is similar to the phosphorylation but different from the TF DBA. Next, we identified TFs that may control the diurnal rhythm of Diurnal rhythmic DBA was closely related with phosphoryla- gene expressions in the mouse liver, and designated these TFs as tion. Phosphorylation plays an important role in TF DNA- “dominant rhythmic TFs” (DR-TFs). We reasoned that DR-TFs 21,22 binding . We next carried out phosphorylation profiling fol- should exhibit diurnal rhythmic DBA, and also control the diurnal 23 lowing the TiO2 enrichment (Supplementary Fig. 4a). We rhythmicity of their downstream target genes. To nominate DR- identified 9465 phosphorylated peptides, representing 1657 TFs, we used the TF and downstream TG (target gene) correlation phosphoproteins in the 16 time points across the diurnal cycles. database from the CellNet25 to integrate the TF DBA and the Phosphorylation level of 89 proteins were judged as diurnal transcriptome data. This analysis allowed us to identify 46 DR-TFs rhythmic (p < 0.1) (Fig. 3a; Supplementary Data 4). Of the 72 TFs out of the 80 diurnal rhythmic TFs (Supplementary Data 8), in the phosphoproteome data set, 55 was identified in the DBA including the well-characterized master clock proteins Bmal1, profiling, and 22 of them were diurnal rhythmic in DBA (p < 0.1), Clock, Dbp, Bhlhe40, and several other TFs that include Stat1, accounting for 40% (22/55) of the phosphorylated TFs. The high Mlxip, Mef2c, Cebpb (Fig. 4d). These DR-TFs may provide a concordance between phosphorylation pattern and DBA of many framework to understand circadian regulation of transcription as TFs was highlighted by the master circadian clock TFs including exemplified in Fig. 4e. For example, the Stat1 and Stat3, which are Bmal1, Dbp, and Clock (Fig. 3b; Supplementary Fig. 4b, Sup- the DR-TFs that peak in the daytime, control the transcription of plementary Data 4). immune-related genes such as Irf5 and Dhx58, which are also We tried to classify the phosphoproteome into four TPs as we peaked in the daytime (Fig. 4e; Supplementary Fig. 8b). observed in the TF DBA, but failed to find clear functional We defined transcriptional “activators” as TFs whose DBA are enrichments. When we separated the phosphorylated proteins positively correlated with transcripts of their target genes into two groups (daytime and nighttime group) based on their (Pearson’s r > 0.5), and “repressors” whose activities are nega- peak time, we found that the daytime group was enriched in tively correlated with transcripts of their target genes (Pearson’s r immune response and lipid metabolism, whereas the nighttime < −0.5). It is interesting to note that DR-TFs peak in a TP tend to group was enriched in the cell cycle and glucose metabolism “activate” TGs enriched in the dominant bioprocess of this TP, (Supplementary Fig. 4c, d, Supplementary Data 5). These results and to “repress” TGs enriched in the dominant bioprocess of suggest that the phosphorylation pattern provides lower time other TPs (Supplementary Fig. 5a, b, Supplementary Data 11). resolution than the TF DBA did in functional enrichments. We then constructed a computational model for the diurnal rhythmic signaling pathways using a recently published data10 and used the Comparison of diurnal rhythm of transcriptome with pro- kinase-substrates relationship database24 to infer the kinases and teome. Previous studies have demonstrated poor correlation used a protein–protein interaction database to find TFs that may between the transcriptome and the proteome16,26,27. We performed be targeted by the kinases. As shown in Fig. 3c, the diurnal whole proteome profiling using the same batch of liver samples rhythmic expressed phosphoproteins turned out to be enriched in (three mice were killed every 3 h for 48 h)28 (Supplementary four bioprocesses: the immune response kinase-TF (K–T) axis Fig. 6a). We identified a total of 6780 proteins, in which 575 pro- mediated by Akt/Pdk1/Erk2–Stat1/Stat3/Nfkb1, and the lipid teins were diurnal oscillating (p < 0.1), accounting for 8% of the metabolism K–T axis mediated by Ampka1/Ck2a1–Nr2f2/Pparδ/ total proteins (Fig. 5a; Supplementary Data 9). The GO bioprocess Vdr at daytime, the cell cycle K–T axis mediated by Pkcd/ enrichment of diurnal rhythmic proteins showed the consistency of Ck1d–Smad4/Smad3 and glucose metabolism K–T axis mediated the dominant diurnal pathways, the gene transcription regulation by Mtor–Rorα/Mef2c/Srebf1 at nighttime. and protein expression in, namely immune response, cell cycle, lipid

Fig. 3 Diurnal phosphoproteome of mouse liver. a Numbers of phosphosites, phosphopeptides, and phosphoproteins detected in our diurnal phosphoproteome of mouse liver. The distribution of diurnal rhythmic phosphoproteins in the total phosphoproteins. Hierarchical clustering of the cycling phosphoproteins ordered by the phase of the oscillation. Values for each phosphoprotein at all analyzed samples (columns) are color code based on the intensities, low (blue) and high (yellow) z-scored normalized iBAQ. The upper white to black bar indicates the 2 days’ cycle. Daytime is shown in white, while nighttime is shown in black. b Temporal abundance of DNA-binding activity of Bmal1, Clock, Dbp, and Nr1d1/Rev-erbα and their correlated abundance of their phosphorylated forms. X axis represents the sampled time points, Y axis represents the z-scored abundance. c Time-resolved map of signaling cascades regulated by diurnal changed phosphoproteins. Proteins (kinases, phosphoproteins, TFs) were colored based on their peak time, daytime-peaked proteins are yellow (detected in this study) or orange (detected in the study by Robles et al.10), while nighttime-peaked proteins were blue (detected in the study by Robles et al.10) or green (detected in this study). Diurnal rhythmic proteins (this study: JTK_CYCLE p < 0.1, Robles et al.10: q (adjusted p) < 0.1) were shown with red border. Box plots show the path from substrates to pathway-specific TFs were shorter than the path from substrates to all detected TFs (pair tailed Student’s t test p < 0.05). For the box plot, the bottom and top of the box are the first and third quartiles, and the band inside the box is the median of the paths between TFs and substrates

6 NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 ARTICLE metabolism, and glucose metabolism (Fig. 5a, b; Supplementary In the proteome data set, we found that, while the mean Fig. 6c), (Supplementary Data 10). Specifically, the core proteins of coefficient variation during the circadian cycle is 0.68, 30% of innate immune response, the key enzymes of the glucose metabo- proteins have coefficient variations larger than 0.9 (Supplemen- lism were dominant during the daytime, while the key enzymes in tary Fig. 6b). We then compared the CVs (coefficient variation) the lipid metabolism and TCA cycle were upregulated during the between proteome and transcriptome, and found that genes nighttime (Supplementary Fig. 6d). whose transcriptome varies greatly than proteome were mainly

a b Diurnal rhythmic transcripts Bmal1 Dbp ZT0 ZT12 ZT24 ZT36 ZT48 3 3 2 2 2 1 1 0 0 -score -score z −2 z –1 –1 –2 –2 Total transcripts 0 12 24 36 48 0 12 24 36 48 11,120 Zeitgeber time (h) Zeitgeber time (h) Diurnal rhythmic Per1 Bhlhe40 3 3 transcripts 2349 2 2 1 1 0 0 -score -score z –1 z –1 –2 –2 0 12 24 36 48 0 12 24 36 48 Zeitgeber time (h) Zeitgeber time (h) c 2 1

-score 0 z ZT0/24 –1 ZT12/36

Lipid metabolism Regulation of cell cycle Cellular lipid metabolism Cell cycle Response to lipid Cell cycle Response to nitrogen compound Regulation of cell cycle 01234 0 1 2 –log (p value) –log (p value) Viral process Response to glucagon Regulation of viral process Glucose metabolic process Immune system process Metabolic process B cell activation Response to insulin 0 1234 0123 –log (p value) –log (p value) Diurnal rhythmic transcripts Diurnal peaked transcripts

d ZT0/ZT24 Gatad1Mafg Creb3l3 Hhex Nfia Sox9 Zhx2 Tfdp1 Stat1 Stat3 Smarcc1 Zhx3 Cebpb Mef2c Rbpj Mlxipl Rfxank Nfatc1 Tbp Rfx1 Smarca5 Smarcc2 Dominant rhythmic Prdm16 Rora TP4 TP1 Ppard -TF 46 8 17 ZT18/ZT42 ZT6/ZT30 TP3 TP2 Diurnal rhythmic-TF 80 3 18

Bhlhe40Dbp Srebf1 TF 297 Npas2 Clock Elk3 Mlxip Tfe3 Pbx3 Prrx1 Rb1 Rfx3 Tcf7l2 ZT12/ZT36 Arntl/Bmal1 Fli1 Foxp1 Sall2 Nfkb1 Nr1d1 Onecut1 Arnt

e Npas2Mlxip Foxp1 Creb3l3 Nfkb1Fli1 Mlxipl HhexStat1 Stat3 Cebpb Zhx3 Srebf1 Prdm16 Nfia

Onecut1 Tcf7l2 Tfe3

Ceacam1Lpin3 G6pdx Gla Stat3 Irgm2Ddx58 Cd79b Samhd1 Nfκbia

Cyp7a1Pmvk Cps1 Serpina6 Elovl3 Fga Nr1d1 Pml Adar Irf5 Ltbr Serpinf2 Sirt3 Cnksr3 Esr1 Stat5b

Baat Acot12 Pla2g12b Il12rb1 Pik3ap1 Kynu Serpinb9

Lipid metabolism Immune response Glucose metabolism Cell cycle

DR-TF whose DNA-binding activity peaked in TP1 DR-TF whose DNA-binding activity peaked in TP3

DR-TF whose DNA-binding activity peaked in TP2 DR-TF whose DNA-binding activity peaked in TP4

NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 enriched in metabolism(Supplementary Fig. 7a). The correlation glycolysis/gluconeogenesis, fatty acid degradation, PPAR signal- coefficients between transcriptome and proteome in all 16 time ing, but the enrichment was not observed in non-ubiquitylated points were between 0.5 and 0.6 (Supplementary Fig. 7b), which diurnal rhythmic proteins (Fig. 5i). Proteins involved in the is consistent with previous reports29,30. Of the 2349 diurnal innate immune response, which include the receptors (Il23r, rhythmic gene transcripts, 1256 of them were identified in the Fcer1g, Fcgr2b et al.), the enzymes and adaptors (Rac1, Trim25, proteome, and only 133 of them displayed diurnal rhythm on Ikk, Trafd1 et al.), and the effector TFs (Stat1 and Stat3) were all both mRNA (p < 0.05) and protein levels (p < 0.1). The diurnal ubiquitylated. The same phenomenon was also found in the fatty rhythmicity of 2216 mRNAs and 442 proteins were detected in acid metabolic process (Fig. 5j). mRNA or protein only. Interestingly, transcript-specific diurnal We also found consistency in the diurnal rhythmicity of rhythmic genes were enriched in RNA processing/splicing, while protein expression and their ubiquitylation level, including protein-specific diurnal rhythmic genes were enriched in toxin Trim25, Iκb, Irgm1 in the immune pathway, Me1, Slc27a5, and transport and lysosomal transport (Fig. 5c). These results suggest Apoa2 in the fatty acid metabolic pathway (Supplementary that the diurnal rhythm of transcriptome is not necessarily Fig. 9). These results collectively suggest that ubiquitylation is translated to the proteome, and the diurnal rhythm can be better another important mechanism in the regulation of diurnal understood by measuring both transcripts and proteins. rhythm in addition to transcription regulation. We also nominated “dominant diurnal rhythmic TFs (DR-TFs)” from the protein profiling data using the similar criteria as we did “ ” The diurnal rhythmic regulatory network of the Med complex. with the transcriptome data set. Surprisingly, 41 DR-TFs were InadditiontoTF,co-regulatorisanotherclassofproteinsthatplay found using the proteome data set, covering 40 of the 46 DR-TFs fi important roles in transcription regulation. The multi-subunit identi ed from the transcriptome data set (Supplementary Fig. 8a, Mediator complex appeared to be a hub of transcription regula- Supplementary Data 8). A substantial overlap of TGs of the DR-TFs tion31. In our TFRE data set, we detected 19 of the 30 components at proteome level were also observed (Supplementary Fig. 8c). of the Med complex and found that many of them exhibited diurnal Together, these results indicated that, while the overall diurnal oscillating pattern. For instance, Med1 and Med27 peaked in the rhythm cannot be directly translated from mRNA to protein, DR- daytime, while Med12 and Med24 peaked in the nighttime (Fig. 6a; TFs derived from the transcriptome and the proteome data sets are Supplementary Data 1).DBAofMed1andMed27showedstrong more consistent, suggesting that TF DBA may be the major driver diurnal rhythmic oscillation (p < 0.1). This observation was con- in governing circadian regulation in mouse liver. firmed by manually extracting XIC (extracted ion chromatogram) and western blotting (Supplementary Fig. 10aandb). An interaction between Med1 and Clock has been reported The dynamic ubiquitylated pattern during the circadian cycle. 32 fi To gain further insight into the circadian clock in mouse liver, we previously . We then calculated the correlation coef cient measured another PTM—ubiquitylation to survey another between components of Med complex with the diurnal rhythmic TFs. The diurnal rhythmic DBA of many TFs, including Bmal1, dimension of the protein landscape during the circadian cycle κ fi (Fig. 5d). We identified 3424 ubiquitylated sites on 1144 proteins Clock, Naps2, Elf2, and Nf- b1 were signi cantly correlated with during one circadian cycle (Fig. 5e; Supplementary Data 12). The that of Med1 (Supplementary Fig. 10c, Supplementary Data 1). It number of the detected ubiquitylated proteins appeared to be is interesting to note that TFs whose DBA peaked in different more in daytime than in nighttime, while the number of total time of the clock tends to correlate with different components of proteins detected in the proteome was similar (Fig. 5e). Clustering the Med complex (Fig. 6b), suggesting that different components analyses also revealed that ubiquitylated proteins that peaked in of the Med complex may be recruited by the diurnal rhythmic the daytime and were enriched in immune response and lipid TFs during the circadian clock. metabolism, while the ubiquitylated proteins that peaked in the We constructed a computational model for a hierarchal network nighttime were enriched in glucose metabolism (Fig. 5f). centered around the Mediator complex (Fig. 6c). We focused on the In the ubiquitylation data set, it appeared that 5% of the four essential diurnal rhythmic biological processes, namely the diurnal rhythmic TFs, particularly DR-TFs, such as Stat1 and immune response, lipid metabolism, cell cycle and glucose Stat3, were ubiquitylated, while 2% of the non-diurnal-rhythmic metabolism, to implicate the role of the Mediator complex in these biological processes. We found that the correlation between the TFs were ubiquitylated. We also compared the ubiquitylation fi patterns of their TGs. We found that 8% of the TGs of the diurnal Med complex and pathway-speci c TFs (TFs enriched in a pathway or bioprocess) is significantly higher than that with all detected TFs rhythmic TFs were ubiquitylated, while 6% of the TGs of the non- ’ − diurnal rhythmic were ubiquitylated (Fig. 5g). Additionally, the (average Pearson s r 0.78 vs 0.19, p value 1.6e 6) (Fig. 6c). Furthermore, the correlations were even higher in the pathway’s number of ubiquitylated TGs of the diurnal rhythmic TFs was fi significantly higher than the number of ubiquitylated TGs of the dominant TP as de ned by TFRE analysis than in the other phases, randomly selected TFs (Fig. 5h), indicating that target gene suggesting that the Med complex serves as a gene regulatory hub in products (GPs) were also further regulated by ubiquitylation, in the diurnal regulation network. addition to their regulation by TFs. The GO analysis revealed that the ubiquitylated diurnal rhythmic proteins were significantly The diurnal switch of the proteome of Kupffer cells. Liver is an enriched in the major diurnal rhythmic pathways, including organ with predominant innate immunity, which can specifically

Fig. 4 Diurnal transcriptome of the mouse liver. a Distribution of diurnal rhythmic transcripts. b Hierarchical clustering of the cycling transcripts ordered by the time of the oscillation. Values for each transcript at all analyzed samples (columns) are color code based on the intensities, low (blue) and high (yellow) z-scored normalized FPKM. The upper white to black bar indicates the 2 days’ cycle. Daytime is shown in white, while nighttime is shown in black. Temporal abundance of transcripts of Bmal1, Dbp, Per1, and Bhlhe40 in 2 consecutive cycles. X axis represents the sampled time points; Y axis represents z-scored FPKM. c Area graph shows the normalized abundance of transcripts that peaked in different time of the day (red: daytime, green: nighttime). Bar plots shows the GO terms enriched by diurnal peaked transcripts (black), and GO terms enriched by diurnal rhythmic transcripts (JTK_CYCLE, p < 0.1) (red). d The hierarchical pyramid shows the number of TFs, diurnal rhythmic TFs, transcriptome nominated DR-TFs. The Rose diagram shows the number of transcriptome nominated DR-TFs of each time phase, TFs were colored based on their peaked phase, (TP1 shows in yellow, TP2 shows in red, TP3 shows in green, TP4 shows in blue). e The regulation network of DR-TFs to their TGs (target genes), TFs were colored based on their peak phase

8 NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 ARTICLE detect infection through pattern-recognition receptors (PRRs) (Irf3, 5, 6, and 9), TFs (Nf-κb2) and the effector cytokines (Il23r, that recognize specific structures, called pathogen-associated Tnfα). Their abundance varied diurnally with peak phase in the molecular patterns (PAMPs)33. Toll-like receptor (TLR) signal- daytime and valley phase in the nighttime. Two Tnfα inducible ing is the dominant PAMPs of the liver34. We detected large proteins, Tnfaip3 and Tnfaip8, were expressed higher in the number of proteins in TLR signaling pathways, including the nighttime, suggesting a temporal regulation of the TLR signaling receptors (Tlr3, Tlr6), the adaptor (Myd88), signal transducers (Supplementary Fig. 6d). The expression level of the Nf-κbib was

a Diurnal GO specific diurnal Total protein rhythmic protein rhythmic protein 6780 ZT0 ZT12 ZT24 ZT36 ZT48 ZT0 ZT12 ZT24 ZT36 ZT48 2 Diurnal rhythmic protein 575 –2 Lipid metabolism Lipid metabolism 61 Immunity Glucose 25 Immunity metabolism 8 Glucose metabolism b c Lipid metabolism Immune response RNA splicing

Inflammatory response RNA processing Lipid metabolism NF-kappaB signaling RNA splicing Innate immune response RNA processing Lipid biosynthesis Mediated immunity 0246 Natural killer cell 051015 –log (p value) –log (p value) 01.520.5 1 –log (p value) Toxin transport Cell cycle Glucose metabolism Ion transmembranetransport Protein targeting to membrane Cell cycle Hexose metabolism Lysosomal transport 01234 Mitotic cell cycle Glucose metabolism –log (p value)

012345 02468 Genes expressed diurnal rhythmically only at mRNA level –log (p value) –log (p value) Diurnal rhythmic proteins Diurnal rhythmic transcripts Genes expressed diurnal rhythmically only at protein level

d IP of peptides LC-MS Protein extraction Trypsin digestion with K-ε-GG antibody K-GG K-GG K-GG K-GG 0ZT6 ZT12 ZT18 K-GG K-GG

e Ubiquitylated sites f ZT0 ZT6 ZT12 ZT18 Ubiquitylated proteins

Protome 6000 Immune response ZT6 ZT12 Total ubiquitylated sites ZT0 137 72 ZT18 4000 5500 3424 69 28 7 5000 211 49 Glucose metabolic process 118 8 4500 Fatty acid metabolic process Total ubiquitylated proteins 312 2000 3000 1144 59 4 2000 Lipid transport 30 19 Ubiquitylated 21 1000

sites & protein number 0 0 ZT0 ZT6 ZT12 ZT18 Whole liver protein number g Carboxylic acid catabolic process 6 10

8 i 8 Ubi-diurnal rhythmic protein 4 Metabolic pathways 6 6 Non-ubi-diurnal rhythmic protein Metabolic pathways Glycolysis / Gluconeogenesis

value) 4 Oxidative phosphorylation 4 p Fatty acid degradation 2 Ubi-R-TF/ R-TF Pentose phosphate pathway Percentage Percentage 2 Ubi-non-R-TF/ non-R-TF 2 PPAR signaling pathway Ubi-R-TF-TG/ R-TF-TG –log ( Ribosome Citrate cycle (TCA cycle) Fatty acid degradation 0 0 Ubi-non-R-TF-TG/ non-R-TF-TG 0 0 5 10 15 20 Fold enrichment

h j Innate immune response Fatty acid oxidation/Fatty acid metabolism process Lipid transport Ub Ub Ub Ub Tlr2 Il23r Fcer1g Apoa2 110 Onecut1 Fcgr2b Ub Ub 100 Ub Fabp Lipogenesis Slc27a2 Ppar Ub 90 Ub Ub Isg15 Ub Ub Ub Ub Scd1 Fads2 80 Tollip Rac1 Stat1 Stat3 Rxr 60 Mal Src Cholesterol Creb3l3 Ub RigI Ub Rnf-125 Cyp7a1

Zhx3 Ub Ub Fatty acid transport Stat1 Ub 40 Mlxip Trim25 Mavs Ub Ub Ub Ub Ifit1 Ub Dbi Fabp1 Acs Sting Ub Srebf1 Ub Fatty acid oxidation Traf3 Trafd1 Ub Nfkb1 Npas2 Ub Ub Number of ubi-TGs Stat3 E1 Cpt-1a Cyp4a12 20 Dbp Ub Nfatc1 ER Mitochondria Prdm16 E2 Rbpj Traf2 Casp8 Ub Gluconeogenesis Pparδ Nr1d1 E3 Ub Cebpb Ub Ub Pepck 0 Ikkε Ikkα Ikkβ Ub Ubiqutylated proteins 3 5 7 19 Ubi-protein peak in the nighttime 12 31 50 79 501 125 199 316 Ubi-protein peak in the daytime Target gene numbers

NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 negatively correlated with that of Nf-κb2 in the circadian clock, aminotransferase) and AST (aspartate aminotransferase) levels consistent with its role as an Nf-κb inhibitor. and histology of the liver. As shown in Fig. 8a, the group treated Since Kupffer cells (KCs), known as liver-resident macro- with LPS at ZT0 had lower survival rate, higher ALT/AST levels, phages, may be the cell-type responsible for the innate immunity and more sever liver damage compared to the group that were in the liver35, we asked whether the daytime and nighttime treated at ZT12. These results suggest that the immune systems in variations of immune response in the liver was attributable to the liver may have differences during the daytime and nighttime variations in KCs. To this end, we isolated primary KCs from and is consistent with reports that temporal differences of – liver at four different time points of the circadian clock (ZT0, macrophages were found in serum, spleen, and lymph nodes38 40. ZT6, ZT12, and ZT18), and performed MS-based proteome We next injected the LPS/D-GalN to the mice in two ZT analysis (Fig. 7a). Interestingly, the number of KCs seemed to points, ZT0 and ZT12, then isolated KCs 2 h after co-injection, vary during the circadian cycle, peaking at ZT6 (1.8e6/liver), and and profiled their proteomes (Fig. 8b). We observed that the falling at ZT12 (4.56e5/liver) (Fig. 7b). Among the 4684 identified abundance of proteins of the innate immune pathway were higher KC proteins, 690 showed diurnal rhythm (p < 0.1) (Fig. 7c; in the daytime than in the nighttime (Fig. 8c). We then focus on Supplementary Data 13); however, the diurnal rhythmicity of the Tlr4 signal transduction pathway, which directly responds to only 61 of them were observed in the whole-liver proteome, LPS. We found that abundance of proteins including Myd88, Nf- suggesting that the cell-type-resolved KC proteome provided κb, Irak4 were highly expressed in the daytime than at nighttime more information than the whole-liver proteome (Fig. 7d). (Fig. 8c), indicating that the Tlr4 signaling transduction pathway Further analyses indicated that as many as 1078 proteins showed were elevated and may be more sensitive to LPS during the strong oscillation between daytime and nighttime with fold daytime. changes >5, although they failed to pass the JTK_CYCLE test. They included many well-characterized immune-related proteins, such as Tmem173 and Mavs. A selected group of proteins with Discussion diurnal expression differences between the KC proteome and the As the major metabolic organ, liver plays an important role in whole-liver proteome were confirmed by MS measurement and regulating metabolism homeostasis around the circadian clock. A western blotting (Supplementary Fig. 11). The GO analysis number of transcriptome and metabolome studies on circadian demonstrated that the diurnal differential expressed proteins in clocks have illustrated the remarkable roles of circadian clock on the KC were much more significantly enriched in the immune cellular and organismal physiology41,42, but relatively little is response pathways than the whole-liver proteome (Fig. 7e). known about the temporal regulation of gene expressions and its Furthermore, the abundance of the major immune response driving force at the proteome level. Two previous studies by pathway appeared to peak in the daytime, as exemplified by the Daniel Mauvoisin et al.15 and Robles et al.16 quantified 5000 and major components of the TLR pathway including Tlr4, Myd88, 3000 proteins, respectively. These studies revealed the dynamics Irak4, and Tak1 (Fig. 7f, g). of diurnal liver proteome. However, due to technical limitations, We further constructed interaction networks between KC and these studies failed to identify the core clock transcription whole-liver proteins around the circadian clock. The connections factors15,16. between KC and whole-liver proteins in immune response In this study, we employed catTFRE12 to profile the dynamics pathway were more dominant in the daytime than the nighttime of TF sub-proteome during circadian cycle. As the result, 80 out (p = 1.5e68), while the opposite phenomenon was observed for of 297 identified TFs were determined as diurnal rhythmic TFs the metabolism process, and more connections between KC (JTK_CYCLE, p < 0.1) during two circadian cycles. Bioinformatic proteins and whole-liver proteins were observed in the nighttime analysis indicated that the diurnal rhythmic TFs could be grouped (p = 1.5e18) than in the daytime (Fig. 7h, i). These results into four major pathways, namely the immune response, glucose suggested a potential cooperation between KC and the liver metabolism, fatty acid metabolism and the cell cycle, based on organ. oscillation patterns of their DBA. We then employed a TLR-induced liver injury model to test The quantification method for the proteome now is mostly whether the diurnal rhythm of innate immune has any biological label-free quantification (LFQ). Compared with label-based MS consequences using LPS treatment36,37. We administrated LPS/ quantification, the LFQ can quantify unlimited number of sam- D-GalN to the mice in two ZT points, ZT0 and ZT12, and then ples for comparison and may also offer higher dynamic range to measured the survival rate, serum ALT (alanine detect low abundance proteins43, the LFQ using the iBAQ

Fig. 5 Diurnal proteome of mouse liver. a Distribution of pathway-specific diurnal rhythmic proteins. Yellow, green, and purple cycle represent the number of proteins enriched in lipid metabolism, immunity, and glucose metabolism. Hierarchical clustering of the total (left) and GO specific diurnal rhythmic proteins (right) ordered by the phase of the oscillation. b The bar plot shows the major GO terms enriched by diurnal rhythmic transcripts (yellow), and diurnal rhythmic proteins (green). c The bar plot shows the GO terms enriched by the genes whose diurnal rhythm were inconsistent in transcript and protein level. d Systematic workflow of the ubiquitylation proteome during the circadian cycle in mouse liver. e The total number of ubiquitylated sites and proteins, and the number of ubiquitylation proteins detected at different ZT points. Bar plot shows the comparison between the numbers ubiquitylated proteins with the number of whole-liver proteins detected at different ZT points. f Hierarchical clustering of the ubiquitylated proteins ordered by the phase of the oscillation. g Bar plots show the percentage ratio between ubiquitylated rhythmic TFs and rhythmic TFs versus the percentage ratio between ubiquitylated non-rhythmic TFs and non-rhythmic TFs (left). The percentage ratio between ubiquitylated rhythmic TFs’ TGs and rhythmic TFs’ TGs versus the percentage ratio between ubiquitylated non-rhythmic TFs’ TGs and non-rhythmic TFs’ TGs (right). h The number of ubiquitylated TGs of the diurnal rhythmic TFs versus the number of ubiquitylated TGs of the randomly selected TFs. i Scatterplot shows statistically enriched GO/KEGG pathways by ubiquitylated diurnal rhythmic proteins versus pathways enriched by non-ubiquitylated diurnal rhythmic proteins. j Systematic overview of signal transduction participated by ubiquitylated proteins. Ubiquitylated proteins were colored based on their peak time. Ubiquitylated proteins peaked in the daytime were shown with red border, Ubiquitylated proteins peaked in the nighttime were shown with green border. For all the Hierarchical clustering heatmap, values for each protein at all analyzed samples (columns) are color code based on the intensities, low (blue) and high (yellow) z-scored normalized iBAQ. The upper white to black bar indicates the diurnal cycle. Daytime is shown in white, while nighttime is shown in black

10 NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 ARTICLE algorism may suffer from lower accuracy and in general was able We aimed to provide multi-dimensional proteomics data sets to quantify a change of more than a factor of 2, which was to portrait the landscape of circadian clock in the mouse liver. adequate for the characterization of many biological processes While centered at TF DBA, our data sets also contain phospho- including circadian rhythm studies published previously10,11. proteome, ubiquitylation proteome, nuclear proteome, whole

a b Mediator complex Med 17 ZT0 ZT12 ZT24 ZT36 ZT48 Med 20 Med 17 Med 30 Med 23 2 Med 19 Med Med 19 Med29 Med 12 Med Med28 Med 8 Med 8 Med 24 Med12 28 Med 6 28 Med 6 Med21 Med 21 Med 29 –2 Med Med20 Med 4 Med30 Med 1 4 Med 1 Med14 Med 7 Med 7 Med23 Med 27 Med 27 Med 14 Med1 Med 14 Med15 Med 15 Med 29 Med 15 Med 16 Med7 Med 16 Med27 Med 20 Med 21 Med 30 Med6 Med 23 Med24 Med Med17 Med 24 12 Med4 Day Night Med16 Med19 Recruit Recruit Med8 Stat3Foxp1 Ppard Stat1 Smad4 Dbp Mef2c Rora

Regulate Regulate Immunity Lipid metabolism Cell cycle Glucose metabolism

c Immune response Lipid metabolism

Med Med Med23 Med7 Med28 Med20 Med4 1.0 Med12 Med27 Med6 1.0 Med1 Med14 0.5 Med15 0.5 Med24 Med24 Med17 Med4 Med29 Med8 Med21 0.0 Med16 0.0 Correlation

Med27 Med21 Correlation Med19 p=1.68e–13 Med19 Med20 p=5.26e–26 Med16 –0.5 –0.5 Med8 Med23 Med12 Pathway All TF Med7 Pathway All TF Med15 specific TF Med29 specific TF Med30 Med30 Med17 Med28 Med1 Med14 Med6 p 0.0075 p 3.91e–10

0.77 0.44 0.79 0.28 TF TF Day Night Shp Rxrg Thrb Fxr Vdr Pparg Day Night Rb1 Nfatc3 Pparg Stat1 Ppard Hnf4g Hnf4a Ear2 Car Esrrg

Stat3 Nfatc1 Foxp1 Nfkb1 Fli1 Lrh1 Tead1 Tfcoup2 Pxr

TG TG C1Qa C1Qc Itgb2 Adh4 G6Pc EhhadhApoa2 Apoa1 Apoa4 Apoc4

Otud7B Gbp6 Samhd1 Gbp2 H2-D1 H2-K1 Angptl3Abcd2 Irs1 Abcb4 Acacb Acox2

Ifit3 C1Ra Aoc3 Cd36 Nfkb2 Anxa3 Cyp2D10 Cyp4F13 Cyp2C70 Cyp2C37 Cyp2J6 Cyp7A1 Pcyt2 Myd88 Cd93 Orm1 Hcls1 Dhx58 Cnn2 Apof Crot B4Galnt1 Ptgis

Cell cycle Glucose metabolism

Med Med Med23 Med29 Med7 Med1 Med23 Med28 Med20 Med15 Med24 Med4 Med14 1.0 1.0 Med24 Med7 Med4 Med17 0.5 Med29 Med30 0.5 Med21 Med15 Med27 Med12 Med19 Med28 0.0 Med16 Med12 0.0 Correlation Med30 Med16 Correlation Med20 Med17 p 1.2e–5 Med21 p 1.05e–6 –0.5 Med6 –0.5 Med1 Med8 Med14 Pathway All TF Pathway All TF Med27 Med8 Med6 Med19 specific TF specific TF p 0.0037 p 5.94e–5 TF TF 0.20 0.70 Smad3 Rora Dbp Arntl –0.30 0.84 Glr Mef2c Srebf1 Rora Day Night Day Night Smad4 Nr1D2 Clock

TG TG Cenpf Cenpe Mcm3 Mcm6 Ranbp1 Topbp1 Pgm3 Pygm Phkg1 Phka1 Pck2 Agl

Tubb5Ccnb2 Mad2L1 Mcm5 Npm1

Peak time p value 0.8 Day p <0.1 Night p <0.1 −0.2

NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications 11 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 proteome, and the whole transcriptome measured from the same DNA length of 2.8 kb. An aliquot of 1 mg NEs (nuclear extractions) were then batch of samples. These data sets would allow other researchers in performed catTFRE pull-down. 2.8 kb catTFRE DNA and biotinylated catTFRE the field to perform in-depth mining of different layers of regulation primers were synthesized by Genscript (Nanjing, Jiangsu Province, China). In total, 3 pmol biotinylated DNA pre-bound to Dynabeads (M280 streptavidin, Thermo of the circadian clock, which should lead to eventual construction of Fisher scientific), and then incubated with NEs in 4 °C for 2 h. After incubation, the the hierarchal circadian rhythmic regulation network. supernatant was discarded, and the NETN buffer (100 mM NaCl, 20 mM Tris-HCl, In summary, we have built a TF DBA-centered multi-dimen- 0.5 mM EDTA, and 0.5% [vol/vol] Nonidet P-40) and PBS was used to wash the μ sional proteomics landscape to illustrate the hierarchal gene Dynabeads beads twice respectively. An aliquot of 20 l of 2× SDS (sodium dodecyl sulfate) loading buffer was used to re-suspend beads. The resuspended beads were expression network in the circadian clock of mouse liver. Our then boiled at 95 °C for 5 min. As for SDS-PAGE separation, samples were stained data sets include TF DBA, protein phosphorylation, ubiquityla- with Coomassie Brilliant Blue, and sliced into six bands equally according to the tion, transcriptome, nuclear protein profiling, and whole pro- molecular weight ranges, followed by in-gel digestion overnight at 37 ℃ with 12 teome profiling as well as the whole proteome of the Kupffer cells. trypsin, referring to the protocol described before . We offered the most comprehensive data sets of the diurnal rhythm in the mouse liver by far and provided the richest data Protein trypsin digestion and first dimension RPLC. For whole-tissue proteome, resource for the understanding of mouse liver physiology around and nuclear sub-proteome, KCs sub-proteome 100 μg whole-tissue proteins were 45 the circadian clock. digested by the FASP procedure . Namely, the protein samples were supple- mented with 1 M dithiothreitol (DTT) to a final concentration of 5 mM and incubated for 30 min at 56 ℃, then added iodoacetamide (IAA) to a 20 mM final Methods concentration, and incubated in the dark at room temperature. After half an hour Animals and tissue collection. Eight-week-old C57BL/6, SPF (specific pathogen- incubation, samples were added 5 mM final concentration of DTT and keep in dark free) male mice (weight 20–25 g) were housed in pathogen-free, temperature- for another 15 min. After these procedures, protein samples were loaded into 10 kD controlled environment, scheduled with 12–12 h light–dark cycles. All mice were Microcon filtration devices (Millipore) and centrifuged at 12,000 × g for 20 min and maintained with free access to food and water for 2 weeks. To collect tissues, three washed twice with Urea lysis buffer (8 M Urea, 100 mM Tris-HCl pH 8.0), twice mice were killed every 3 h for 48 h. We grouped animals randomly, to ensure that with 50 mM NH4HCO3. Then the samples were digested using trypsin at an the investigator was blinded to the group allocation during the experiment. This enzyme to protein mass ratio of 1:25 overnight at 37 °C. Peptides were extracted study is under the guidelines of the animal care regulations of Fudan University, and dried (SpeedVac, Eppendorf). and received ethical and scientific approval from Fudan University. As for the whole-liver proteome, the digested peptide then performed first dimension RPLC before LC-MS/MS. The dried peptides were loaded into a homemade Durashell RP column (2 mg packing (3 μm, 150 Å, Agela) in a 200 μl Kupffer cell isolation and protein extraction. Normal eight-week-old C57BL/6 tip), then eluted sequentially with nine gradient elution buffer which contains – – male mice were housed, and scheduled with 12 12 h light dark cycles for 2 weeks. mobile phases A (2% acetonitrile (ACN), adjusted pH to 10.0 using NH .H O) and ’ 3 2 At each time point three mice were used for KCs isolation. We used two-step liver 6%, 9%, 12%, 15%, 18%, 21%, 24%, 30%, 35% mobile phase B (98% ACN, adjusted 44 perfusion digestion in situ followed the previously described protocol . KCs were pH to 10.0 using NH .H O). The nine fractions then were combined into six fi 3 2 rst isolated simultaneously with LSECs using collagenase-based density gradient groups (6% + 24%, 9% + 30%, 12% + 35%, 15%, 18%, 21%) and dried under TM centrifugation. Cells between 11.2% and 17% in optiprep density gradient were vacuum for sub-sequential MS analysis. carefully collected. Then cells were labeled with phenotypic markers (FITC-con- jected CD11b and PE-conjected F4/80) and purified with FACS analysis (BD Biosciences, Franklin Lakes, NJ). Collected KCs were then quickly frozen with Phosphoproteome sample preparation. A total of 1 mg whole-liver protein lysates − liquid Nitrogen and transferred to 80 °C refrigerator for storage. were digested with trypsin for the TiO2 enrichment. To the digested peptides, 0.25 ml binding buffer (80% ACN, 5% trifluoroacetic acid (TFA (Sigma-Aldrich)), and 1 M Protein sample preparation. Three mice livers collected at each zeitgeber time lactic acid (Sigma-Aldrich)) was added, peptides were mixed at room temperature for point were pooled together to extract nuclear protein and whole-tissue protein. 1 min at 2000 rpm, cleared by centrifugation, and transferred to a clean 0.5 ml tube. Nuclear extractions: 0.3 g pooled tissues were homogenized with 800 μl TiO2 beads were subsequently added to peptides at a ratio of 4:1 beads/protein, and Cytoplasmic Extraction Reagent I (CER I) buffer (NE-PER™ kit Thermo Fisher incubated at room temperature for 30 min on rotor with middle speed. Beads were Scientific), and 8 μl protease inhibitor (Pierce™, Thermo Fisher Scientific). Nuclear subsequently pelleted by centrifugation for 2 min at 2000 × g, and the supernatant proteins were extracted following the protocol provided by the manufacturer. (containing non-phosphopeptides) was collected to repeat the enrichment for twice Protein concentrations were measured using Bradford method (Eppendorf and then discarded. Beads were suspended in wash buffer (80% ACN and 5% TFA), transferred to a clean 0.5 ml tube, and washed a further four times with 0.5 ml wash Biospectrometer). fi Whole-tissue protein extractions: An aliquot of 0.1 g pooled tissues were lysed buffer. After the nal wash, beads were suspended in 0.05 ml wash buffer, three batch with 400 μl urea lysis buffer (8 M urea, 100 mM Tris-HCl pH 8.0), 4 μl protease beads were all transferred onto a 0.2 ml StageTip with two pieces of C8, and cen- inhibitor (Pierce™, Thermo Fisher Scientific) was added to protect protein from trifuged for 3 min at 500 × g or until no liquid remained on the StageTip. Bound degradation and protein concentrations were measured using Bradford method phosphopeptides were eluted two times with different gradient of elution buffer (0%, 3%, 6%, 9%, 12%, 40% ACN, and 15% NH4OH) and collected by centrifugation into (Eppendorf Biospectrometer). + + + KCs protein extractions: In total, 100 μl cell pellets were lysed with 400 μl urea six 0.5 ml tube, then combined them into three tubes (0% 40%, 3% 12%, 6% lysis buffer (8 M urea, 100 mM Tris-HCl pH 8.0), 4 μl protease inhibitor (Pierce™, 9%). Peptide samples were concentrated in a SpeedVac. Peptides were resuspended in Thermo Fisher Scientific) was added to protect protein from degradation and buffer containing 5% MeOH and 10% FA for liquid chromatography-tandem mass protein concentrations were measured using Bradford method (Eppendorf spectrometry (LC-MS/MS) analysis. Biospectrometer). LC-MS/MS analysis. Peptides from catTFRE tandem in-gel digestion were CatTFRE pull-down and trypsin digestion. We referred to TF-binding database detected by Q Exactive Plus (Thermo Fisher Scientific) and peptides used for JASPAR to select consensus TFREs for different TF families. To design the detecting proteome, nuclear proteome, phosphoproteome, and ubiquitylation catTFRE construct, we used 100 selected TFREs and placed two tandem copies of proteome and KC sub-proteome were detected by Orbitrap Fusion Lumos each sequence with a spacer of three nucleotides in between, resulting in a total (Thermo Fisher Scientific).

Fig. 6 The diurnal rhythm of the Mediator complex. a Hierarchical clustering of the proteins ordered by the phase of the oscillation. Values for each protein at all analyzed samples (columns) are color code based on the intensities, low (blue) and high (yellow) z-scored normalized iBAQ. The upper white to black bar indicates the 2 days’ cycle. Daytime is shown in white, while nighttime is shown in black. b Modular structure of Mediator complex recruited by diurnal rhythmic TFs that regulated different bioprocess, subunits colored in yellow or green represent subunits’ high or low DNA-binding activity, respectively. c The regulation network of Mediators, TFs, and TGs in four major pathways. For each pathway, the network on the left shows the TFs recruited components of Mediator to regulate TGs’ expression, proteins (TFs, components of Mediator, TGs) were colored based on their abundance (proteins peaked in the daytime are yellow, in the nighttime are green), and diurnal rhythmic ones were shown with red border (JTK_CYCLE p < 0.1). The upright box plot shows the correlation between pathway-specific TFs with Mediators is higher than the average correlation between Mediators with total TFs (pair tailed Student’s t test p < 0.05). For the box plot, the bottom and top of the box are the first and third quartiles, and the band inside the box is the median of the correlation between TFs and Mediators. The Heat map on the down-right shows the diurnal difference of the correlation between pathway-specific TFs with Mediators (pair tailed Student’s t test p < 0.05)

12 NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 ARTICLE

Q Exactive Plus LC-MS/MS analyses were performed on an Easy-nLC 1000 μm, 120 Å, Dr. Maisch GmbH) with a linear 5–35% Mobile Phase B (ACN and liquid chromatography system (Thermo Fisher Scientific) coupled to an Q-Exactive 0.1% formic acid) at 600 nl/min for 75 min. The MS analysis was performed in a Plus via a nano-electrospray ion source (Thermo Fisher Scientific). The peptides data-dependent manner with full scans (m/z 300–1400) acquired using an Orbitrap from in-gel trypsin digestion were dissolved with 10 μl loading buffer (5% mass analyzer at a mass resolution of 70,000 at m/z 400. The top twenty precursor methanol and 0.2% formic acid), and 5 μl was loaded onto a 360 μm I.D. × 2 cm, ions were selected for fragmentation in the HCD cell at normalized collision energy C18 trap column at a maximum pressure 280 bar with 12 μl solvent A (0.1% formic of 30%, and then fragment ions were transferred into the Orbitrap analyzer acid in water). Peptides were separated on 150 μm I.D. × 14 cm column (C18, 1.9 operating at a resolution of 17,500 at m/z 400. The automatic gain control (AGC)

a b Kupffer cells 2.5e6

2e6

1.5e6 0 ZT12 ZT24 1e6

LC-MS 5e5 Trypsin digestion Protein extraction Kupffer cell number 0 0 ZT6 ZT12 ZT18 Zeitgeber time (h)

c Diurnal rhythmic KC proteins deDiurnal rhythmic KC protein Diurnal rhythmic whole live protein KC proteins 7 5000 6 Lipid metabolic process Fatty acid metabolic process 4000 Cellular lipid metabolic process KC Apoptotic process Steroid metabolic process 3000 proteins whole liver proteins 4 Response to hormone Lipid biosynthetic process 2000 p value<0.1 61 p value<0.1

value) Regulation of T cell 1000 629 514 p Leukocyte proliferation

Protein number 2 Response to interleukin-4

0 –log ( 0 6 12 18 Response to cytokine Regulation of adaptive immune response Zeitgeber time (h) 0 012345 Fold enrichment

fgImmune system Regulation of innate h Toll-like receptor signaling process immune response Immune related proteins peaked in daytime Tlr4 p 3.9e–19 p 4.3e–5 Immune related proteins peaked in nighttime 8 4000 750 p value 1.5e–68 Myd88 2000 Tlr3 Tlr8 6 311 1500 Myd88 4 FOT*e6

FOT*e6 Irak4 1000 831 500 2

0 1 Average interaction Endosome Trafd1 Rip1 ZT0/6 ZT12/18 ZT0/6 ZT12/18 0

Innate immune response RigI Lipid related proteins peaked in daytime Lipid related proteins peaked in nighttime p 0.0003 Nemo Ikkα Tak1 6 p value 1.5e–18 4000 Traf3

4 1500 Tbk1 Nfκbib FOT*e6 1000 Cytoplasm 2 1.5 500 Irf3 Nfκb

0 Average interaction 0 Nucleus −1 0 ZT0/6 ZT12/18 i

Innate immune response Blood coagulation, leukocyte Response to homone Lipid metabolism Innate immune response Endosomal transport Ion transport Eukocyte activation Ifitm3 Serpinb1d Fatty acid metabolic process C3 Stat6 Ehd1 Mtmr2 Tfrc Mertk Ifitm2 Aacs Slc9a3r1 Camp Ifit3 Slc25a17

Ep300 Agt Srebf1 Fyb Yes1 Slc27a5 Snx27 Was Thy1 Fgr Crat Tmem165 Map3k7 Leukocyte migration Ikbkg Them4 Zfyve16 Aldh5a1 Akt2 S100a8 S100a9 Ripk3 Junb Thnsl2 Irs1 Ddx17 Kras Cd180 Myd88 Slc27a1 Mtor Oxsm Fcer1g Tgfb1 Mapk3 Nfkb1 Fcgr4 Pik3r1 Tlr7 Tlr4 Hadh Fcgr1 Nos3 Sirt4 Parp1 Blood coagulation Crot Tlr8 Tlr3 Mapkapk2 Tek Safb Cdk2 Cdk5 Map4k2 Tlr6 Ramp2 Rps6ka1 Nasp Toll-like receptor signaling pathway Rela Smarcc1 Tnxb Ldlr Lpin1 Hnf4a Tmprss6 Trf Pcna Vps41 Prpf8 Regulation of JUN kinase activity Lpl Cav1 Nedd4l Ptk2b Response to hormone Serpina1cSerpina1e Apoa1 Apoa2 Amfr Alms1 Cdkn1b Cpb2 Lpin2 C9 Tnf Rnf126 Ube2d2a F5 Response to lipid F13a1 Regulation of cytokine secretion Cholesterol homeostasis F13b F12 Ube2r2Ube2h Pros1 Serpind1 Diurnal rhythmic whole liver protein (p value <0.1) Protein K48-linked ubiquitination F2 Fgg Diurnal rhythmic KC protein (p value <0.1) Serpinf2 Diurnal changed KC protein Complement and coagulation cascades KC protein Whole liver protein

NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications 13 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 for full MS was set to 3e6, and that for MS/MS was set to 5e4, with maximum ion The peptides were analyzed using full scan plus PRM modes. The full mass within injection times of 20 and 60 ms, respectively. The dynamic exclusion of previously the range of 300 to 1400 m/z was collected. The MS1 resolution was set at 120,000 acquired precursor ions was enabled at 18 s. (at 200 m/z), MS2 methods were controlled with a timed inclusion list containing Orbitrap Fusion Lumos LC-MS/MS analyses were performed on an Easy-nLC the target precursor m/z value, charge, and a 3 min retention time window that was 1000 liquid chromatography system (Thermo Fisher Scientific) coupled to an determined from DDA results. All of the raw files were processed using Skyline 3.1. Orbitrap Fusion Lumos via a nano-electrospray ion source (Thermo Fisher Scientific). The intensities of three fragment ions were summed for peptide quantification. The Fractions from the first dimension RPLC were dissolved with loading buffer (5% intensities of up to three peptides were summed and used for GP quantitative methanol and 0.2% formic acid), and loaded onto a 360 μm I.D. × 2 cm, C18 trap comparison. column at a maximum pressure 280 bar with 12 μl solvent A (0.1% formic acid in water). Peptides were separated on 150 μm I.D. × 14 cm column (C18, 1.9 μm, 120 Å, RNA-Seq. About 0.1 g tissues were pooled together and extract total RNA using Dr. Maisch GmbH) with a series of adjusted linear gradients according to the fi hydrophobicity of fractions with a flow rate of 600 nl/min. The MS analysis was Trizol regent (Thermo Fisher Scienti c). Total RNAs were subsequently enriched performed in a data-dependent manner with full scans (m/z 300–1400) acquired for polyadenylated RNA with the Oligotex mRNA Mini Kit (QIAGEN). For using an Orbitrap mass analyzer at a mass resolution of 120,000 at m/z 200. The top sequencing, we used Illumina HiSeqX platform with Pair End 150 reads, and speed data-dependent mode was selected for fragmentation in the HCD cell at sequence at depth averaging 50 million reads per sample. The results were aligned normalized collision energy of 32%, and then fragment ions were transferred into the to mouse genome (UCSC version mm10) using tophat2 (v2.0.12), and transcript ion trap analyzer with the AGC target at 5e3 and maximum injection time at 35 ms. abundance was calculated using cuffnorm (v2.2.1), showed as FPKM (fragments The dynamic exclusion of previously acquired precursor ions was enabled at 12 s. per kilobase of exon model per million mapped reads) value. The FPKM of each sample were calculated z-score using the equation z = (x −μ)/σ,(μ stands for the mean of the samples’ FPKM of one cycle, and σ stands for Label-free based MS quantification for proteins. Raw MS files were processed the standard deviation of the samples’ FPKM of one cycle). with the MaxQuant software (version 1.5.3.30), using the integrated Andromeda search engine with FDR < 1% at peptide and protein level. As a forward database, TF classification. Proteins identified by DBA profiling were categorized into DBPs, the mouse refseq protein database (2013.07.01) was used. A reverse database for the TFs, and TCs. We extracted DBPs by filtering the genes’ description “DNA-binding”. decoy search was generated automatically in MaxQuant. Enzyme specificity was set fi ‘ ’ And we extracted TFs and TCs by ltering the gene symbols, using the gene symbols to Trypsin , and a minimum number of seven amino acids were required for list of TFs and TCs, from public databases described in previous studies14. peptide identification. For label-free protein quantification, the ‘Match Between Runs’ option were used with a matching window of 3 min to transfer MS1 iden- tification between Runs. For raw MS files from catTFRE sub-proteome, proteome Bioinformatics and statistical analysis. To determine the subset of cycling TFs, and nuclear proteome, default settings were used for variable and fixed modifica- proteins, and transcripts. We performed JTK_CYCLE test with period range: tions (variable modification: acetylation (Protein-N terminus) and oxidation 20–28 h and the amplitude and phase as free parameters. A statistical cut-off of p < methionine (M), fixed modification: carbamidomethylation (C)). For raw MS files 0.1 was used to define the cycling proteome, and p < 0.05 was used to define the from phosphoproteome, setting for variable and fixed modifications included cycling transcripts. Hierarchical clustering was performed using the pheatmap variable modifications for oxidation (methionine), acetylation (protein N-term), (Pretty Heatmaps) function in the R package (pheatmap, version 1.0.8). The GO and phos-pho (STY) and fixed modifications for carbamidomethyl (C). For raw MS terms that were enriched in the sets of enriched genes were determined using the files from ubiquitylation proteome, setting for variable and fixed modifications Database for Annotation, Visualization and Integrated Discovery (DAVID) included variable modifications for oxidation (methionine), acetylation (protein N- Bioinformatics Resource v 6.7 with Fisher’s exact test. term), and GlyGly (K) and fixed modifications for carbamidomethyl (C). We used MaxQuant LFQ algorithm46 to quantitate the MS signals, and the proteins’ Bioinformatic analysis of kinase signaling pathway. We extracted pathway- intensities were represented in iBAQ. fi fi fi The iBAQ30 (intensity-based absolute protein quantification) of each sample speci c phosphorylated proteins and pathway-speci cTFsby ltering their GO terms. were transferred into FOT (a fraction of total protein iBAQ amount per The hierarchical network of kinase, phosphoproteome, and TFs were constructed = −μ σ μ based on multiple layers. The network among kinase and substrates were annotated experiment), and calculated z-score using the equation z (x )/ ,( stands for 24 the mean of the samples’ FOT of one cycle, and σ stands for the standard deviation using information accessed from PhosphoSitePlus , and the protein interaction net- of the samples’ FOT of one cycle). work among TFs and phosphorylated proteins were generated with the STRING v10.0 using medium confidence (0.4) and experiments and database as active interaction sources. The network was visualized using Cytoscape v 3.3.0. The shortest path from Parallel reaction monitoring analysis for KC. Data-dependent acquisition. Pep- pathway-specific phosphorylated proteins to TFs were based on the protein interaction tides were eluted from a 150 μm I.D. × 2 cm, C18 trap column and separated on a data from STRING, and calculated using Rscript: shortest.paths.r. homemade 150 μm I.D. × 30 cm column (C18, 1.9 μm, 120 Å, Dr. Maisch GmbH) – with a 150 min non-linear 5 35% ACN gradient at 600 nl/min. The combined Bioinformatic analysis of Mediator-centered network. We extracted pathway- method consisted of an MS1 scan at a resolution of 120,000 (at 200 m/z) with an fi fi – speci c TFs and TGs by ltering their GO terms. The hierarchical network of AGC value of 3e6, max injection time of 80 ms and scan range from m/z 300 1400, Mediators, TFs and TGs were constructed based on multiple layers. The protein top30 precursor ions from MS1 were selected for MS2 scans with higher-energy fi = interaction network among Mediators were performed with the STRING v10.0 collision dissociation detected in the Orbitrap rst (R 15,000 at 200 m/z, AGC using medium confidence (0.4) and “experiments and database” as active inter- target 2e4, max injection time 20 ms, isolation window 1.6 m/z, normalized colli- action sources. The network between Mediators and pathway-specific TFs were sion energy of 27%, The dynamic exclusion of previously acquired precursor ions based on their Pearson’s correlation coefficient, we linked each pathway-specificTF was enabled at 25 s). ’ fi fi to three Mediators with highest Pearson s correlation coef cient with it. The net- Peptides of target GPs were identi ed and focused in the exact RT windows by work among TFs and TGs was obtained from CellNet25, Data visualization was data-dependent acquisition (DDA) scan (Supplementary Data 14). The parent ions done with Cytoscape v 3.3.0. in the table were monitored on an Easy-nLC system (Thermo Fisher Scientific, USA) coupled with Q-Exactive HF (Thermo Fisher Scientific, USA). Peptides were separated on a homemade 150 μm I.D. × 30 cm column (C18, 1.9 μm, 120 Å, Dr. Bioinformatic analysis of KC and whole-liver protein network. We extracted Maisch GmbH) with a 150 min non-linear 5–35% ACN gradient at 600 nl/min. pathway-specific proteins by filtering their GO terms. The interaction between KC

Fig. 7 The diurnal switch of the immune response in mouse liver. a Systematic workflow of the diurnal KC sub-proteome. b Diurnal oscillation of the number of KCs during the circadian cycle. c Bar plot shows distribution of the diurnal oscillated proteins of KCs at each ZT point. d Venn plot shows the overlap among diurnal rhythmic whole-liver proteins (JTK_CYCLE p < 0.1) and diurnal rhythmic KC proteins (JTK_CYCLE p < 0.1). e Scatterplot shows statistically enriched GO/KEGG pathways by diurnal rhythmic KC proteins and diurnal rhythmic whole-liver proteins. f Box plots shows the pathway- specific KC proteins’ abundance of expression at ZT0 versus their expression at ZT12 (pair tailed Student’s t test p < 0.05). For the box plot, the bottom and top of the box are the first and third quartiles, and the band inside the box is the median of the proteins’ abundance. g Schematic representation of diurnal changed TLR pathway-related proteins; for each protein, the corresponding expression level was represented by color. The color bar indicates normalized z-scored iBAQ. h Box plots show the stronger connection (more interactions) among immune-related KC proteins and whole-liver proteins in the daytime and stronger connection (more interactions) among lipid-related KC proteins and whole-liver proteins in the nighttime (pair tailed Student’s t test p < 0.05). For the box plot, the bottom and top of the box are the first and third quartiles, and the band inside the box is the median of the interactions between KC proteins and whole-liver proteins. i Diurnal interaction network among KC proteins and whole-liver proteins. KC upregulated proteins were colored in green, whole-liver upregulated proteins were colored in red. Diurnal rhythmic proteins (whole-liver proteins: JTK_CYCLE p < 0.1, KC proteins: JTK_CYCLE p < 0.1) were shown with red border, diurnal changed KC proteins (fold change > 5) were shown with black border

14 NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2 ARTICLE

a LPS

ZT0 Batch #1 Batch #2 1.0 ZT12 ALT AST 0.8 p value=7.91e–8 p value=2.59e–9

0.6 10,000 8000 ZT0

0.4 8000 Survival 6000 0.2 6000 4000 0 p value < 0.05 4000 ALT(IU/L) AST(IU/L) 2000 1 6121824 2000 Hours after LPS injection ZT12 0 0 ZT0 ZT12 ZT0 ZT12

b LPS LPS

Protein extractionTrypsin digestion LC-MS 0 ZT12 ZT24

Kupffer cells

c RIG-I-like receptor Toll-like receptor mTOR signaling pathway signaling pathway signaling pathway TNF signaling pathway 300 120 14 p value 0.04 40 p value 0.006 50 p value 0.05 p value 0.019 12 30 40 40 10 30 30 8 20 6 20 FOT*e6 FOT*e6 FOT*e6 FOT*e6 20 4 10 10 10 2 0 0 0 0 ZT0 ZT12 ZT0 ZT12 ZT0 ZT12 ZT0 ZT12

**** * ** LPS-treated mice at ZT0 2.0 5 15 4 8 1.5 4 3 6 3 10 LPS-treated mice at ZT12 1.0 2 4 2 5

FOT*e6 FOT*e6 FOT*e6 FOT*e6 * p value <0.05 ** p value <0.01 0.5 1 FOT*E6 1 2 0.0 0 0 0 0 ZT0 ZT12 ZT0 ZT12 ZT0 ZT12 ZT0 ZT12 ZT0 ZT12 Toll-like receptor signaling Irak4 Ikbkα Nfκb Irf3 Myd88

Fig. 8 Liver’s diurnal changed immune response to LPS. a Survival curves of mice after co-injected with LPS/D-GalN at ZT0 (red line) or ZT12 (green line) (n = 6 per group). Bar plots shows serum ALT, AST level of mice 6 h after been co-injected with LPS/D-GalN, at ZT0, or at ZT12. Data are mean SD (n = 6–8). Pair tailed Student’s t test was performed. Mice liver sections was prepared 6 h after LPS/D-GalN co-injection at ZT0 or at ZT12, stained with haematoxylin and eosin. The arrows indicated the focal necrosis area (scale bars = 100 µm). b Systematic workflow of the KCs’ sub-proteome after LPS/D- GalN administration in circadian clock of mouse liver. c Box plots shows immune-related pathways were stronger activated by LPS/D-GalN at ZT0 than at ZT12. Bar plots shows higher expression of the TLR pathway-related proteins at 2 h after LPS/D-GalN administration at ZT0 or than at ZT12 (pair tailed Student’s t test p < 0.05). For the box plot, the bottom and top of the box are the first and third quartiles, and the band inside the box is the median of the proteins’ abundance proteins and whole-liver proteins were calculated based on information from Received: 3 July 2017 Accepted: 20 March 2018 protein–protein interaction database STRING (v10.0, medium confidence (0.4) and “experiments and database” as active interaction sources).

Western blotting. Denatured nuclear extractions or proteins were loaded onto polyacrylamide gels, and after electrophoresis, proteins were transferred onto PVDF membranes (Millipore), then blocked in 5% non-fat milk for 1 h at 25 °C. References Incubation of primary antibodies was done overnight at 4 °C in 5% non-fat milk on 1. Saini, C., Suter, D. M., Liani, A., Gos, P. & Schibler, U. The mammalian TBS with 0.1% Tween-20. Primary antibodies were as follows: Med1 (CRSP1/ circadian timing system: synchronization of peripheral clocks. Cold Spring TRAP220) (Bethyl, Cat.: A300-793A, 1:1000), Bmal1 (Cell Signaling Technology, Harb. Symp. Quant. Biol. 76,39–47 (2011). Cat.: 14020 S, 1:1000), Rora (Proteintech, Cat.: 10616-1-AP, 1000), Pparδ (Pro- 2. Reppert, S. M. & Weaver, D. R. Molecular analysis of mammalian circadian teintech, Cat.: 60193-1-Ig, Clone No.: 1B10E1, 1:1000), Tmem173 (Proteintech, rhythms. Annu. Rev. Physiol. 63, 647–676 (2001). Cat.: 19851-1-AP, 1:1000), Nfκb2 (Proteintech, Cat.: 10409-2-AP, 1:1000). The 3. Takahashi, J. S., Hong, H. K., Ko, C. H. & McDearmon, E. L. The genetics of uncropped scans of blots of Bmal1, Rora, Pparδ, Med1, Tmem173, and Nfκb2 were mammalian circadian order and disorder: implications for physiology and represented in Supplementary Fig. 12. disease. Nat. Rev. Genet 9, 764–775 (2008). 4. Cho, H. et al. Regulation of circadian behaviour and metabolism by REV- – Data availability. All Mass Spectrum raw data and the MaxQuant output tables ERB-alpha and REV-ERB-beta. Nature 485, 123 127 (2012). have been deposited to iProX and can be accessed with the iProX accession: 5. Takeda, Y. et al. Retinoic acid-related orphan receptor gamma (RORgamma): IPX0001145000, (TF DBA pattern and whole-liver proteome), IPX0001158000, a novel participant in the diurnal regulation of hepatic gluconeogenesis and (Nuclear proteome, phosphoproteome, ubiquitylation proteome and KC sub-pro- insulin sensitivity. PLoS Genet. 10, e1004331 (2014). teome). RNA-seq data have been deposited to Sequence Read Archive (SRA), with accession number: SRP133633.

NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications 15 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-018-03898-2

6. Shostak, A. et al. MYC/MIZ1-dependent gene repression inversely coordinates 37. Hou, X., Zhou, R., Wei, H., Sun, R. & Tian, Z. NKG2D-retinoic acid early the circadian clock with cell cycle and proliferation. Nat. Commun. 7, 11807 inducible-1 recognition between natural killer cells and Kupffer cells in a novel (2016). murine natural killer cell-dependent fulminant hepatitis. Hepatology 49, 7. Bass, J. & Takahashi, J. S. Circadian integration of metabolism and energetics. 940–949 (2009). Science 330, 1349–1354 (2010). 38. Keller, M. et al. A circadian clock in macrophages controls inflammatory 8. Hendriks, I. A., Treffers, L. W., Verlaan-de Vries, M., Olsen, J. V. & Vertegaal, immune responses. Proc. Natl Acad. Sci. USA 106, 21407–21412 (2009). A. C. SUMO-2 orchestrates chromatin modifiers in response to DNA damage. 39. Marpegan, L. et al. Diurnal variation in endotoxin-induced mortality in mice: Cell Rep. https://doi.org/10.1016/j.celrep.2015.02.033 (2015). correlation with proinflammatory factors. Chronobiol. Int. 26,1430–1442 (2009). 9. Aebersold, R. & Mann, M. Mass spectrometry-based proteomics. Nature 422, 40. Nguyen, K. D. et al. Circadian gene Bmal1 regulates diurnal oscillations of 198–207 (2003). Ly6C(hi) inflammatory monocytes. Science 341, 1483–1488 (2013). 10. Robles, M. S., Humphrey, S. J. & Mann, M. Phosphorylation is a central 41. Panda, S. et al. Coordinated transcription of key pathways in the mouse by the mechanism for circadian control of metabolism and physiology. Cell. Metab. circadian clock. Cell 109, 307–320 (2002). 25, 118–127 (2016). 42. Eckel-Mahan, K. L. et al. Coordination of the transcriptome and metabolome 11. Wang, J. et al. Nuclear proteomics uncovers diurnal regulatory landscapes in by the circadian clock. Proc. Natl Acad. Sci. USA 109, 5541–5546 (2012). mouse liver. Cell. Metab. 25, 102–117 (2017). 43. Xie, F., Liu, T., Qian, W. J., Petyuk, V. A. & Smith, R. D. Liquid 12. Ding, C. et al. Proteome-wide profiling of activated transcription factors with a chromatography-mass spectrometry-based quantitative proteomics. J. Biol. concatenated tandem array of transcription factor response elements. Proc. Chem. 286, 25443–25449 (2011). Natl Acad. Sci. USA 110, 6771–6776 (2013). 44. Ding, C., Wei, H., Sun, R., Zhang, J. & Tian, Z. Hepatocytes proteomic 13. Hughes, M. E., Hogenesch, J. B. & Kornacker, K. JTK_CYCLE: an efficient alteration and seroproteome analysis of HBV-transgenic mice. Proteomics 9, nonparametric algorithm for detecting rhythmic components in genome-scale 87–105 (2009). data sets. J. Biol. Rhythms 25, 372–380 (2010). 45. Wisniewski, J. R., Zougman, A., Nagaraj, N. & Mann, M. Universal sample 14. Ravasi, T. et al. An atlas of combinatorial transcriptional regulation in mouse preparation method for proteome analysis. Nat. Methods 6, 359–362 (2009). and man. Cell 140, 744–752 (2010). 46. Cox, J. et al. Accurate proteome-wide label-free quantification by delayed 15. Mauvoisin, D. et al. Circadian clock-dependent and -independent rhythmic normalization and maximal peptide ratio extraction, termed MaxLFQ. Mol. proteomes implement distinct diurnal functions in mouse liver. Proc. Natl Cell Proteom. 13, 2513–2526 (2014). Acad. Sci. USA 111, 167–172 (2014). 16. Robles, M. S., Cox, J. & Mann, M. In-vivo quantitative proteomics reveals a key contribution of post-transcriptional mechanisms to the circadian Acknowledgements regulation of liver metabolism. PLoS Genet. 10, e1004047 (2014). This work is supported by National Key R&D Program of China (2017YFA0505102), 17. Chawla, A., Repa, J. J., Evans, R. M. & Mangelsdorf, D. J. Nuclear receptors National Natural Science Foundation of China (31770886, 31770892 and 31700682), and lipid physiology: opening the X-files. Science 294, 1866–1870 (2001). Chinese Ministry of Science and Technology (2016YFA0502500), National Key R&D 18. Yang, X. et al. Nuclear receptor expression links the circadian clock to Program of China (2017YFC0908404), National Program on Key Basic Research Project metabolism. Cell 126, 801–810 (2006). (973 Program, 2014CBA02000), National Institute of Health (Illuminating Druggable 19. Xiao, L., Xie, X. & Zhai, Y. Functional crosstalk of CAR-LXR and ROR-LXR in Genome, Grant U01MH105026) and Grant for Shanghai Municipal Science and Tech- drug metabolism and lipid metabolism. Adv. Drug Deliv. Rev. 62, 1316–1321 nology Major Project (2017SHZDZX01), International Science & Technology Coopera- (2010). tion Program (2012DFB30080). 20. Mitro, N. et al. The nuclear receptor LXR is a glucose sensor. Nature 445, 219–223 (2007). Author contributions 21. Lamia, K. A. et al. AMPK regulates the circadian clock by cryptochrome C.D., J.Q., F.C.H. conceived the project and designed the experiment. Y.Z.W., L.S., M.W. phosphorylation and degradation. Science 326, 437–440 (2009). L., R.Y.L., R.G., J.B.Q. conducted the experiments. Y.Z.W., L.S, W.L.L, Q.Z. performed 22. Chiu, J. C., Ko, H. W. & Edery, I. NEMO/NLK phosphorylates PERIOD to data analysis. Y.Z.W., C.D. wrote the manuscript. Y.W., J.Q., F.C.H., B.Z. edited the initiate a time-delay phosphorylation circuit that sets circadian clock speed. manuscript. Cell 145, 357–370 (2011). 23. Mazanek, M. et al. Titanium dioxide as a chemo-affinity solid phase in offline phosphopeptide chromatography prior to HPLC-MS/MS analysis. Nat. Additional information Protoc. 2, 1059–1069 (2007). Supplementary Information accompanies this paper at https://doi.org/10.1038/s41467- 24. Hornbeck, P. V. et al. PhosphoSitePlus: a comprehensive resource for 018-03898-2. investigating the structure and function of experimentally determined post- translational modifications in man and mouse. Nucleic Acids Res. 40, Competing interests: The authors declare no competing interests. D261–D270 (2012). 25. Cahan, P. et al. CellNet: network biology applied to stem cell engineering. Cell Reprints and permission information is available online at http://npg.nature.com/ 158, 903–915 (2014). reprintsandpermissions/ 26. Ghazalpour, A. et al. Comparative analysis of proteome and transcriptome variation in mouse. PLoS Genet. 7, e1001393 (2011). Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in 27. Gygi, S. P., Rochon, Y., Franza, B. R. & Aebersold, R. Correlation between published maps and institutional affiliations. protein and mRNA abundance in yeast. Mol. Cell. Biol. 19, 1720–1730 (1999). 28. Ding, C. et al. A fast workflow for identification and quantification of proteomes. Mol. Cell. Proteom. 12, 2370–2380 (2013). 29. Nagaraj, N. et al. Deep proteome and transcriptome mapping of a human Open Access This article is licensed under a Creative Commons cancer cell line. Mol. Syst. Biol. 7, 548 (2011). Attribution 4.0 International License, which permits use, sharing, 30. Schwanhausser, B. et al. Global quantification of mammalian gene expression adaptation, distribution and reproduction in any medium or format, as long as you give control. Nature 473, 337–342 (2011). appropriate credit to the original author(s) and the source, provide a link to the Creative 31. Allen, B. L. & Taatjes, D. J. The Mediator complex: a central integrator of Commons license, and indicate if changes were made. The images or other third party transcription. Nat. Rev. Mol. Cell. Biol. 16, 155–166 (2015). material in this article are included in the article’s Creative Commons license, unless 32. Lande-Diner, L., Boyault, C., Kim, J. Y. & Weitz, C. J. A positive feedback loop indicated otherwise in a credit line to the material. If material is not included in the links circadian clock factor CLOCK-BMAL1 to the basic transcriptional article’s Creative Commons license and your intended use is not permitted by statutory – machinery. Proc. Natl Acad. Sci. USA 110, 16021 16026 (2013). regulation or exceeds the permitted use, you will need to obtain permission directly from 33. Gao, B., Jeong, W. I. & Tian, Z. Liver: an organ with predominant innate the copyright holder. To view a copy of this license, visit http://creativecommons.org/ – immunity. Hepatology 47, 729 736 (2008). licenses/by/4.0/. 34. Akira, S. & Takeda, K. Toll-like receptor signalling. Nat. Rev. Immunol. 4, 499–511 (2004). 35. Li, P., He, K., Li, J., Liu, Z. & Gong, J. The role of Kupffer cells in hepatic © The Author(s) 2018 diseases. Mol. Immunol. 85, 222–229 (2017). 36. Jiang, W., Sun, R., Wei, H. & Tian, Z. Toll-like receptor 3 ligand attenuates LPS-induced liver injury by down-regulation of toll-like receptor 4 expression on macrophages. Proc. Natl Acad. Sci. USA 102, 17077–17082 (2005).

16 NATURE COMMUNICATIONS | (2018) 9:1553 | DOI: 10.1038/s41467-018-03898-2 | www.nature.com/naturecommunications