Uncovering the Proteome Response of the Master Circadian Clock to Light Using an AutoProteome System

Abstract

In mammals, the suprachiasmatic nucleus (SCN) is the central circadian pacemaker that governs rhythmic fluctuations in behavior and physiology in a 24-hr cycle and synchronizes them to the external environment by daily resetting in response to light. The bilateral SCN is comprised of a mere ∼20,000 neurons serving as cellular oscillators, a fact that has, until now, hindered the systematic study of the SCN on a global proteome level. Here we developed a fully automated and integrated proteomics platform, termed AutoProteome system, for an in-depth analysis of the light-responsive proteome of the murine SCN. All requisite steps for a large-scale proteomic study, including preconcentration, buffer exchanging, reduction, alkylation, digestion and online two-dimensional liquid chromatography-tandem MS analysis, are performed automatically on a standard liquid chromatography-MS system. As low as 2 ng of model protein bovine serum albumin and up to 20 μg and 200 μg of SCN proteins can be readily processed and analyzed by this system. From the SCN tissue of a single mouse, we were able to confidently identify 2131 proteins, of which 387 were light-regulated based on a spectral counts quantification approach. Bioinformatics analysis of the light-inducible proteins reveals their diverse distribution in different canonical pathways and their heavy connection in 19 protein interaction networks. The AutoProteome system identified vasopressin-neurophysin 2-copeptin and casein kinase 1 delta, both of which had been previously implicated in clock timing processes, as light-inducible proteins in the SCN. Ras-specific guanine nucleotide-releasing factor 1, ubiquitin protein ligase E3A, and X-linked ubiquitin specific protease 9, none of which had previously been implicated in SCN clock timing processes, were also identified in this study as light-inducible proteins. The AutoProteome system opens a new avenue to systematically explore the proteome-wide events that occur in the SCN, either in response to light or other stimuli, or as a consequence of its intrinsic pacemaker capacity.

Through evolution, the circadian timekeeping system has arisen to ensure that an organism can anticipate and adapt to regular environmental changes resulting from a 24-hr day/night cycle. Virtually all aspects of mammalian physiology and behavior are governed by the circadian clock and exhibit daily rhythms. Intense research and interest in understanding clock timing mechanisms is fueled by the recent discoveries that disruptions in circadian rhythms are linked to a host of pathophysiological disorders, including cancer, cardiovascular disease, metabolic syndrome, and various neurological syndromes (1–3). Genetic studies have established that circadian rhythms are driven by a network of core clock components that interact within a series of dynamically regulated transcription-translation feedback loops (4–6). As the master circadian pacemaker in mammals, the suprachiasmatic nucleus (SCN)¹ can run autonomously with near 24-hr periodicity and coordinate the phase of peripheral oscillators throughout the body. Moreover, the SCN resets its phase in direct response to environment light, ensuring that clock-controlled processes remain tied to the rhythms of the environment. Large-scale gene expression analyses of the murine SCN have revealed hundreds of cyclic transcripts (7, 8). For a handful of these genes, rhythms in protein expression or post-translational modification have also been documented (9, 10). In comparison, proteome-wide analysis of the SCN in response to light has been limited (11, 12), and studies on light-induced protein expression or post-translational modifications are generally restricted to several of the core clock genes (e.g. Period1 and Period2), immediate early genes (e.g. c-Fos), kinases and histones (13–15). A systematic study of the proteome-wide alterations taking place in the SCN throughout the circadian cycle or as a function of light stimulation would further our understanding of the clock timing machinery.

Mass spectrometry-based proteomics is the method of choice for the high-throughput identification and quantification of proteins on a global level (16–19). Generally, relatively large amounts of biological materials (e.g. milligrams to grams of tissues) are needed for proteomics analysis. Unfortunately, in some instances the amount of starting material is limited, as is the case of the murine SCN, a bilateral structure in the hypothalamus that measures ∼0.3 mm³ and is comprised of ∼20,000 neurons functioning as cellular oscillators as well as glial cells. Nonetheless, a recent proteomics study based on difference gel electrophoresis two-dimensional-DIGE successfully visualized 871 protein “spots” from the murine SCN and identified 34 circadian regulated proteins by preparative scale gel and liquid chromatography-tandem MS (LC-MS/MS) (20). Endogenous peptides secreted from the SCN were also studied by both in situ solid-phase extraction followed by off-line matrix-assisted laser desorption ionization/time of flight (MALDI-TOF) MS as well as LC-Fourier transform (FT) MS using multiple tissues (21, 22).

One of the main technical challenges when dealing with minute amounts of starting material is sample loss and lower yield. In recent years, great success has been made to develop integrated fluidic systems for online LC-MS/MS analysis, which provided a clear improvement in sensitivity (23–25). Moreover, immobilized enzymes coupled with high performance liquid chromatography (HPLC) were reported to improve the processing of proteomic samples prior to mass spectrometry (26–30). However, these sample processing systems are not readily compatible with large-scale proteomics studies of minute sample amounts. The automated LC-MS/MS systems are often complex systems, and do not incorporate protein preconcentration and most of the biochemical and chemical processing for proteomic samples. Furthermore, two-dimensional liquid chromatography fractionation, which is necessary for large-scale proteomics study, is still performed with difficulty on current automated systems.

Here, we report a fully automated online proteomics sample processing and mass spectrometry identification system, termed AutoProteome system, for minute amounts of protein samples. All requisite steps for proteomics sample processing, including protein preconcentration, buffer exchange, reduction/alkylation, and digestion are integrated. Furthermore, the system includes online two-dimensional chromatography separation and mass spectrometry identification (two-dimensional LC-MS/MS) based on a standard nanoflow LC-MS/MS system. We used this automated system to study the effect of light stimulation on protein expression in the murine SCN. In total, 196,109 peptides corresponding to 4189 unique proteins were identified by the AutoProteome system with FPR of 1%, and 2131 proteins were quantified with spectral counts-based quantitation, a marked improvement over previous studies. Of these 4189 unique proteins, 387 proteins have statistically significant changes in expression levels (either up- or down-regulated) in response to light. Systematic function analysis revealed their tight connection and distribution in multiple pathways related to circadian rhythms. Subsequent validation by immunostaining and Western blot analysis demonstrated the light-inducible expression of two key proteins that have previously been implicated in clock timing processes, as well as three additional proteins that have not been linked to circadian rhythms.

EXPERIMENTAL PROCEDURES

Materials

Tris(2-carboxyethyl) phosphine hydrochloride (TCEP) was purchased from Pierce (Rockford, IL). Trypsin was purchased from Promega (Madison, WI). Protease inhibitors and phosphatase inhibitor mixture tablets were obtained from Roche (Mannheim, Germany). PolySULFOETHYL A was obtained from (PolyLC, Columbia, MD). POROS HS 20 μm with dynamic binding capacity of 75 mg/ml for lysozyme at pH 6.2 was obtained from Applied Biosystems (Foster City, CA), and Magic C18 AQ (5 μm, 200 Å pore) was obtained from Michrom BioResources (Auburn, CA). Fused-silica capillaries with ID of 200 μm and 700 μm were obtained from Polymicro (Phoenix, AZ). Acetonitrile with 0.1% formic acid and water with 0.1% formic acid were purchased from J.T. Baker (Phillipsburg, NJ). All other chemicals were obtained from Sigma-Aldrich (Saint Louis, MO).

Light Treatment Paradigm and Tissue Processing

Adult (8- to 10-week-old) C57Bl6/J male mice (The Jackson Laboratory, Bar Harbor, ME) were stably entrained to a 12 h light:12 h dark (LD) cycle (100 lux intensity) for a minimum of 2 weeks, and then transferred to total darkness for 2 consecutive days. Following dark adaptation, mice received a single light pulse (LP: 15 min, 100 lux) at circadian time (CT) 15, returned to darkness and killed 4 h later for tissue harvesting. Mice were killed by cervical dislocation and decapitated, and eyes were covered with black electrical tape under dim red light. For immunofluorescence (IF) labeling, brains were dissected and immersed in chilled oxygenated physiological saline, cut into 600 μm coronal sections with an oscillating tissue slicer, and fixed (6 h, room temperature) in a 4% (w/v) formaldehyde/phosphate-buffered saline (PBS) solution, pH 7.4. Tissue was cryoprotected in 30% sucrose (w/v) overnight, and then cut into thin (40 μm) sections using a freezing microtome. For Western blotting and MS, SCN tissue was dissected using a micropunch, frozen on dry ice, and stored at −80 °C until use. Dark control (DD) mice were killed at the same CT without exposure to a single light pulse. For these experiments, CT was estimated based on the prior LD cycle, where Zeitgeber time (ZT) 12 (i.e. lights off) was used to define CT 12. Mice were maintained at the animal facility of the University of Ottawa in accordance with institutional guidelines. All animal handling and experimental procedures were approved by the Animal Welfare Committee of the University of Ottawa.

Immunofluorescent Labeling and Image Acquisition

Thin (40 μm) brain sections were washed five times in PBST (PBS with 0.1% Triton X-100), blocked (1 h, room temperature) in 10% horse serum/PBST, and incubated (overnight, 4 °C) in primary antibodies. The following primary antibodies were used: mouse monoclonal anti-glial fibrillary acidic protein (GFAP) (1:1000; Molecular Probes, Eugene, OR), guinea pig polyclonal anti-arginine vasopressin (AVP) (1:4000; Peninsula Laboratories, San Carlos, CA); rabbit polyclonal anti-RASGRF1 (1:500; Cell Signaling Technology, Danvers, MA); rabbit polyclonal anti-UBE3A (1:3000; LifeSpan Biosciences, Seattle, WA); rabbit polyclonal anti-USP9X (1:500; Novus Biologicals, Littleton, CO); and rabbit polyclonal anti-DsRED (1:200; Clontech, Mountain View, CA). The following day, sections were washed five times in PBST and incubated (2 h, room temperature) with Alexa Fluor® 488- and/or Alexa Fluor® 594-conjugated secondary antibodies (1:1000; Molecular Probes) against the IgG domains of the primary antibodies. In some experiments, sections were counterstained with the nuclear marker DRAQ5™ (1:2000; Biostatus Ltd., Shepshed, Leicestershire, UK). Sections were mounted on slides with Dako Fluorescence Mounting Medium (Dako Canada, Inc., Mississauga, Ontario, Canada).

Images were captured using a Zeiss 510 laser scanning confocal microscope (Carl Zeiss MicroImaging GmbH, Göttingen, Germany) with argon (488 nm), helium/neon (546 nm), and helium/neon (633 nm) lasers. Images were acquired with ZEN 2008 software (Carl Zeiss MicroImaging GmbH). All confocal parameters (pinhole, contrast, brightness, etc.) were held constant for all data sets from the same experiment. For quantitative analysis, central SCN images were inverted and the SCN was outlined using a polygon. The “measure ” function in ImageJ (http://rsbweb.nih.gov/ij) yielded a “mean gray ” value for the SCN. Values were presented as mean relative abundance of the protein examined.

Western Blot Analysis

SCN tissue was pooled from three animals per condition, and homogenized in ice-cold radioimmunoprecipitation assay buffer supplemented with protease inhibitor mixture, incubated for 20 min on ice, and passed several times through a 27-gauge syringe needle. After preclearing by centrifugation (15 min at 17,000 × g), proteins were quantified by the Bradford method using the Coomassie Plus™ Protein Assay Reagent (Fisher Scientific Canada, Whitby, ON, Canada). Protein samples were resolved on sodium dodecyl sulfate polyacrylamide gels under denaturing conditions, and blotted onto polyvinylidene difluoride membranes (Immobilon-P; Millipore) by wet transfer for 2 h at 80 V for low molecular weight proteins or overnight (∼16 h) at 30 V for high molecular weight proteins. Membranes were blocked (45 min, room temperature) with 5% skim milk in TBST (Tris-buffered saline containing 0.05% Triton X-100), and incubated (4 °C, overnight) with the following antibodies: rabbit anti-casein kinase Iδ (CSNK1D) (1:2500, Santa Cruz Biotechnology, Inc., Santa Cruz, CA), rabbit anti-Ras-GRF1 (1:2000, Cell Signaling Technology) and rabbit anti-actin (1:30,000, Sigma-Aldrich). Membranes were incubated (2 h, room temperature) with ImmunoPure® HRP-conjugated goat anti-rabbit IgG (H+L) secondary antibody (1:250,000; Pierce, Rockford, IL). Membranes were washed 5 × 5 min in TBST after antibody incubations, and the signal was detected by chemiluminescence using the SuperSignal® West Femto Maximum Sensitivity Substrate (Pierce). All western quantitation were performed with ImageJ software (http://rsbweb.nih.gov/ij/).

Protein Extract Preparation

For tissue lysis for MS, one SCN tissue was suspended in 100 μl lysis buffer (10 mm HEPES, pH 7.4, 150 mm NaCl, 1% Triton X-100, 2 mm CaCl₂, 2 mm MgCl₂, and fresh protease inhibitor mixture), and sonicated four times for 3 s each with at least 1 min on ice between each pulse. After incubating for 30 min on ice, the extract was clarified by centrifugation at 13,000 × g for 10 min (4 °C). The supernatant was transferred to a fresh tube and the protein concentration was determined by the Bradford method.

Automated Online Proteomics System (AutoProteome System)

As shown in Fig. 1, the AutoProteome system consists of an Agilent 1200 series capillary pump, an Agilent 1200 series micro-autosampler, and a LTQ linear ion trap mass spectrometer equipped with a nanospray source (Thermo, San Jose, CA). The spray voltage was set at 1.8 kV. All MS and MS/MS spectra were acquired in the data-dependent mode. The mass spectrometer was set such that one full MS scan was followed by ten MS/MS scans for LTQ linear ion trap mass spectrometer. Different dimensions of reactor columns (i.e. 200 μm ID × 2 cm, 200 μm ID × 20 cm and 700 μm ID × 10 cm) packed with POROS SCX beads in-house were connected in Agilent 1200 autosampler. Both C18 trap columns (200 μm ID × 5 cm) and C18 analytical columns (75 μm ID × 5 cm) were packed with Magic C18 AQ beads and connected in mass spectrometry part. All of the protein samples and reagents were kept on 96-well sample plates in autosampler at 4 °C in the dark. The flow rate was set at 10 μl/min with 0.1% formic acid for automated sample processing operation, and switched to 5 μl/min to reach a final flow rate of 200 nL/min after splitting for LC-MS/MS analysis.

Fig. 1.

Schematic overview of the automated online proteomics sample processing and analysis of SCN tissue harvested from a single mouse. Micrograph of a coronal mouse brain section containing the SCN (orange box).

All automated proteomic reactor operations were programmed and controlled by the HPLC system and mass spectrometry software in coordination (as shown in supplemental Table S1). The reactor column was first conditioned with 0.1% formic acid by capillary pump, and system pressure was kept below 20 bar. Protein samples were acidified to ∼pH 3 with 100 mm citric acid and loaded onto the reactor column. Trypsin was dissolved into 10 mm citric acid, pH 3 and loaded onto the reactor column in a ratio of 1:5 to protein sample. After the sample and trypsin were loaded, the reactor column was washed with 20% acetonitrile, 10 mm citric acid, pH 3 to remove buffers and contaminants. For protein reduction, 20 mm TCEP, pH ∼2.7 was loaded onto the reactor column. Once the column was saturated with TCEP solution, the loading was stopped by switching the AS valve to position A, and the column was incubated for 10 min. The reactor column was then washed with water and dried by injecting air. Once the reactor column was dried, 100 mm Tris-HCl, pH 8 with 10 mm iodoacetamide was loaded onto the reactor column for alkylation and digestion at the same time and incubated for 60 min by using the same operation for reduction. When the trypsin digestion was finished, the MS valve was switched to position A, and the digested peptides were eluted from the reactor column to the C18 trap column by injecting increasing concentrations of NH₄OAc, pH 2.7. Finally, the AS valve was kept at position A and reverse phase-based gradient elution was applied by the pump to introduce peptides into the mass spectrometer for identification.

Data Analysis

Peak lists of the raw files were generated by Mascot Distiller (version 2.0.0.0, Matrix Science, London, UK). For standard protein analysis, the acquired MS/MS spectra were searched against a database containing 35 standard proteins. For SCN proteome analysis, the acquired MS/MS spectra were searched against the forward and reverse mouse International Protrin Index protein sequence database (version 3.52, 110,604 protein entries; European Bioinformatics Institute, www.ebi.ac.uk/IPI/). The database search was performed using Mascot 2.2.02 (Matrix Science) with the following parameters: carbamidomethyl (C) was set as a fixed modification of 57.0215 Da, whereas oxidation (M, +15.99492 Da) was set as a variable modification. The precursor and fragment mass tolerances were set at 2.0 and 0.8 Da, respectively. Mascot cutoff score was set to 30. The number of allowed missed cleavages was set to 2 and the significance threshold set to 0.05 (p < 0.05). Protein hit required at least one “bold red peptide.” The false positive rate was controlled to be less than 1% using the equation FPR = # false peptides/(# true peptides + # false peptides) * 100 (31). The LTQ raw data associated with this study may be downloaded from www.oisb.ca/downloads.htm.

Spectral counts-based quantitation was carried out using PROTEOIQ software (version 1.5.02, BIOINQUIRE, Athens, GA). Four biological replicates were performed. Identified proteins were kept with at least two unique peptides and from three out of four biological replicates. Furthermore, all the unique peptides could only be assigned to one protein group, and identified protein from one protein group with highest sequence coverage was reported. Spectral counts were normalized across four biological replicates and two biological groups. Significant difference were evaluated by the Fisher's exact test as reported previously,(32) and the identified proteins with p value lower than 0.05 were kept for further function analysis.

Protein function, pathway, and network of the identified proteins were mapped and summarized by Ingenuity Pathways Analysis (IPA), version 8.5 (Ingenuity Systems, Redwood City, CA). Canonical pathways analyses were performed with p value of 0.05 (Fisher's exact test). Networks were displayed with minimum significant score of 16 (Fisher's exact test).

RESULTS AND DISSCUSION

Automated Online Sample Processing and Analysis

We developed an automated system, termed the AutoProteome system, for the processing of minute levels of proteomic samples. The AutoProteome system fully integrates the extraction and concentration of proteins, the chemical and biochemical processing of proteins, and the separation of peptides by two-dimensional HPLC. The AutoProteome system consists of a modified proteomic reactor(33) incorporated in a standard microflow LC-MS setup using only one pump, one autosampler, and one mass spectrometer (Fig. 1), thereby greatly minimizing the complexity of the system. Briefly, in this arrangement the proteomic reactor is introduced right after the autosampler arm prior to the switching valve in the autosampler, thus allowing the whole protein processing procedure to be performed automatically and controlled by the HPLC system software. The autosampler in combination with the HPLC pump can automatically introduce the proteomic samples, the enzymes, and the chemicals necessary for the processing of the proteomic samples. Moreover, the proteomic reactor simultaneously serves as a preconcentration device for the proteins. The protocol for the proteomic reactor (33) was modified such that dithiotreitol was replaced with TCEP on the proteomic reactor, allowing the protein reduction to be accomplished under acidic conditions, and some of the buffer exchange procedures used in the manual proteomic reactor protocol were eliminated. As shown in supplemental Fig. S1, the reduction reaction efficiency is not affected by the replacement of dithiotreitol with TCEP, whereas the online sample processing procedure is greatly simplified. The processing of proteomic samples on the AutoProteome system, including the concentration, and biochemical and chemical processing of the samples, is completed within 120 to 170 min, depending on the sample volume.

The AutoProteome system also performs online two-dimensional LC-MS/MS analysis. SCX based peptide separation has been extensively used as the first separation mode in two-dimensional LC-MS/MS analysis (34). Interestingly, the proteomic reactor is based on a strong cation exchange mechanism: therefore, once the processing of the samples has occurred on the proteomic reactor, it can then be used as the first dimension separation column in two-dimensional LC-MS/MS analysis. This is accomplished by sequential step-elution of the peptides from the proteomic reactor to a C18 precolumn (Fig. 1 and supplemental Table S1).

We first compared the performance of the AutoProteome system with the manual proteomic reactor. Different amounts of BSA were processed on the AutoProteome system with online one dimension LC-MS/MS and the manual proteomic reactor with offline one dimension LC-MS/MS. As shown in Fig. 2A, better performance was obtained on the AutoProteome system for the analysis performed using 20 ng to 2 μg of protein. The most probable reason for this improvement is the reduction in sample loss in the fully automated system.

Fig. 2.

The performance of the AutoProteome system for BSA protein standard and SCN tissue from single mouse. A, Detection limit of the AutoProteome system and manual proteomic reactor for the BSA protein standard from 2 ng to 2 μg (n = 3); B, The distribution of identified proteins and identified peptides from different sample amounts and two-dimensional LC-MS/MS methods (n = 2). For the analysis of 20 μg of SCN tissue extract, a reactor column with dimensions of 200 μm ID × 20 cm was used, and NH₄OAc with concentrations of 50, 100, 300, 700, and 1000 mm was used. For the analysis of 200 μg of SCN tissue extract, a reactor column with dimensions of 700 μm ID × 20 cm was used, and NH₄OAc with concentrations of 50, 100, 150, 200, 250, 300, 350, 400, 700, and 1000 mm were used for 10 SCX fractions elution and 25, 50, 75, 100, 125, 150, 175, 200, 225, 250, 275, 300, 325, 350, 375, 400, 500, 600, 700, and 1000 mm were applied for 20 SCX fraction elution. C, The distribution of the unique peptides identified from 200 μg of SCN tissue extract in 10 SCX fractions, and (D) their overlap between different fractions.

We then tested the performance of the system for the analysis of the complex SCN proteome. In particular, we tested whether a single or a fraction of a mouse SCN would provide sufficient amounts of starting material for analysis on the AutoProteome system. We also tested whether different numbers of step elution from the proteomic reactor would augment the number of identified proteins. We tested the performance of the system for 20 and 200 μg of protein extract of SCN tissue (corresponding to 1/10 and 1 mouse SCN, respectively) and for 5- to 20-step elutions during the two-dimensional peptide separation. The number of proteins identified from the SCN using the AutoProteome system depends on the amount of samples injected on the system and the depth of the two-dimensional-HPLC separation (Fig. 2B). Moreover, a doubling in the number of identified peptides was obtained when the step elution was increased from 10 to 20. However, the number of unique identified proteins did not significantly change, and therefore, we decided to utilize 200 μg of sample and perform 10-step elutions per analysis for subsequent analyses. The total time for the AutoProteome process including the sample processing and the two-dimensional LC-MS/MS was 28 h. Based on this strategy as shown in Fig. 2C, approximately on average each fraction is composed of 76% unique peptides and two subsequent fractions are composed of 90% unique peptides. Furthermore, the heat map analysis of the identified unique peptides across different fractions confirmed their tight distribution at single fraction (Fig. 2D). These results demonstrate the efficient fractionation of the digested peptides and the possibility of using the AutoProteome system for the analysis of the SCN proteome.

Characterization of the Proteomic Response of the SCN to Light

As the master circadian pacemaker in mammals, the SCN coordinates peripheral clocks throughout the body in order to maintain proper phase alignment in relation to the external environment. The SCN is innervated by glutamatergic and PACAPergic projections arising from the retina, and can therefore directly reset (or entrain) its clock phase in response to light, the most potent synchronizer or “Zeitgeber” (“Zeit” = time, “geber” = giver) of circadian rhythms. Photic stimulation during the circadian night activates a number of kinase pathways (e.g. p42/p44 MAPK) within the SCN that ultimately trigger de novo transcription of several immediate early genes (e.g. c-fos, fos-B, jun-B) (35), core clock genes (e.g. PER1 and PER2) (14, 36), and other light-inducible genes (12). Studies have shown that transcriptional activation within the SCN is essential for light-induced resetting of behavioral, physiological, and metabolic rhythms (37). In addition, various studies have demonstrated a requirement for de novo protein synthesis in light-evoked clock entrainment (38, 39). More recently, we have shown that light-induced microRNA expression in the SCN acts as a feedback modulator of clock entrainment via regulation of protein translation (40, 41). To better understand the molecular processes that underlie the ability of the SCN to synchronize to environmental light, we set forth to identify the proteins that are altered in expression in the murine SCN in response to photic stimulation.

To this end, C57Bl6/J mice that had been stably entrained to a fixed 12 h:12 h LD cycle were maintained in constant darkness for 2 consecutive days prior to receiving a 15-min light pulse (LP) at CT 15. Following light exposure mice were returned to dark conditions and killed 4 h later for harvesting of SCN tissue. Dark control (DD) mice were killed at the same CT without receiving the LP. SCN tissues from separate cohorts of animals were used for immunohistochemical analysis, Western blotting and proteomic analysis. The protein extracts from the SCN of LP- and DD-treated mice were sequentially analyzed on the AutoProteome system. In order to explore the light-responsive proteome of the SCN, a quantitative proteomic study based on spectral counts was performed following photic stimulation (42–44). For each treatment condition (i.e. LP and DD), four biological replicates of 200 μg proteins (i.e. ∼1 bilateral SCN) were analyzed on the AutoProteome system. For each individual protein, relative quantitation of its abundance between the light- versus dark-exposed mice was obtained. As a control to confirm the efficiency of our light treatment paradigm, we verified by indirect immunofluorescence that the mPeriod 2 (mPER2) gene, a known light-induced clock gene in the SCN (14), was up-regulated in the SCN of LP-treated mice relative to DD subjects (Fig. 3A).

Fig. 3.

Spectral counts-based quantitative study of light-responsive proteome of murine SCN. A, Light-induced expression of the mPER2 gene in the SCN tissue as determined with mPER2::DsRED transcriptional reporter mice (67). mPER2::DsRED mice received a brief light pulse (LP: 100 lux, 15 min) at CT 15 and were sacrificed 4 h later. Dark control mice (DD) were not exposed to light but were killed at the same CT. DsRED expression (red) was visualized by indirect immunofluorescence using a DsRED-specific antibody. B, Number of proteins identified from DD and LP samples; C, MA plot showing the protein expression level distribution of the identified proteins. Proteins identified with significant up-regulation (p value < 0.05) are labeled with red, and proteins identified with significant down-regulation (p value < 0.05) are labeled with green.

A total of 196,109 peptides corresponding to 4189 unique proteins were identified using the AutoProteome system with a FPR of 1%. In order to meet our very stringent filtering criteria established to decrease the false positive rate, each individual protein must be identified: (1) in at least three out of four biological replicates; and (2) with a minimum of two unique peptides. We identified 1958 proteins from the LP sample with FPR of 0.18%, and 1716 proteins from the DD sample with FPR of 0.11% (supplemental Table S2). Four hundred fifteen (19.5%) and 173 proteins (8.2%) were uniquely identified in the LP and DD samples, respectively (Fig. 3B).

We evaluated the reproducibility of the spectral count method and our filtering criteria for four biological replicates of the light and dark treated animals (as shown in supplemental Fig. S3A and B). The normalized spectral counts of each protein between any of two biological replicates were correlated and compared, and the correlation values were calculated. The average correlations between replicates for the LP and DD samples are 0.96 and 0.95, respectively. These results demonstrate that the AutoProteome system, in combination with the spectral count method and our filtering criteria, has good reproducibility for the semiquantification of proteins.

We were particularly interested in the proteins that had statistically significant changes in their expression levels following light exposure. We established an MA plot based on the log2 transformed normalized spectral counts of each identified proteins where M (log₂LP-log₂DD) represents the differential protein abundance between DD and LP samples, and A ((log₂LP+log₂DD)/2) represents the average protein abundance. Proteins that do not vary in expression following a light pulse are expected to center at zero on the M axis. As shown in Fig. 3C, the majority of the proteins quantified were tightly centered at zero. All of the proteins present on the MA plot were color-coded according to the p value calculated by Fisher's exact test as previously described (32). Proteins with a p value > 0.05 were coded blue, whereas proteins with a p value < 0.05 were coded red or green if their expression levels were enhanced or decreased, respectively, following the light treatment. From our analysis, 218 proteins were significantly up-regulated (supplemental Table S3), and 169 proteins were significantly down-regulated (Table S4). Interestingly, the minimum spectral counts for up-regulated proteins and down-regulated proteins were 6 and 4, respectively, reinforcing the notion of our statistic filtering approach.

We next explored the possible overlap between rhythmically expressed genes and light-inducible genes in the SCN. To this end, we compiled a list of 416 cycling transcripts from two microarray gene expression profiling studies of the murine SCN (7, 8). This list of rhythmically expressed genes was compared with our list of proteins whose expression was affected by light treatment. Of the 416 previously reported rhythmically expressed genes, 120 were detected by the AutoProteome system in this study. Out of these 120 proteins, 23 exhibited a significant change at the protein level following photic stimulation (supplemental Table S5). This is a reasonable correlation between the circadian transcriptome and the light-responsive proteome considering that they are not fully overlapping processes and that the changes at the protein levels are often out-of-phase or not directly correlated with the changes at the transcript level.

Interestingly, of the 23 proteins that were significantly altered by light treatment and exhibited circadian rhythms in gene expression, five proteins that are implicated in the ubiquitin-proteosome pathway were found: four of these (i.e. PSMD5, UBA3, UBQLN1, and USP9X) were significantly up-regulated, whereas one (i.e. FBXO3) showed light-dependent down-regulation. Protein ubiquitination has been previously associated with the degradation of several core clock genes, subsequently influencing the accumulation of the protein and, hence, the pace of circadian transcription and circadian oscillations (45). For instance, PER1 and PER2 are targeted to the ubiquitin-proteasome degradation pathway by the actions of casein kinase 1 epsilon (CK1δ) and βTrCP, an F-box protein that comprises the SCF (Skp1-Cul1-F-box protein) ubiquitin ligase (46, 47). Loss of FBXL3, another F-box protein that targets the ubiquitination and degradation of the core clock genes cryptochome 1 (Cry1) and cryptochrome (Cry2), slows the circadian clock (48, 49). With respect to our findings, PSMD5 belongs to the 26S proteasome complex and functions as one of the non-ATPases. UBA3 (ubiquitin-activating enzyme 3) was shown to mediate protein degradation through the 26S proteasome (50). UBQLN1 (ubiquilin 1) physically interacts with both proteasomes and ubiquitin ligases in large complexes, and is postulated to regulate protein degradation (51). FBXO3 (F box only protein 3) was recently identified as a new component of the SCF ubiquitin ligase complex (52). The clustering of these ubiquitination-related proteins demonstrates their potential involvement in regulating light-induced clock entrainment.

In addition to factors implicated in the ubiquitin-proteasome pathway, molecular functions of our light-responsive proteome data and the circadian transcriptome data both show enrichment in functions such as transporter, transcription regulator, kinase, and oxidoreductase, demonstrating potential overlap between circadian genes and light-inducible genes in these molecular function groups (supplemental Fig. S3). Moreover, we compared our results with the recent study of the rhythmically expressed proteins in the SCN (20). Of the 34 circadian proteins previously identified by two-dimensional-DIGE coupled with MALDI-TOF MS, 26 proteins were also identified by the AutoProteome system, with three proteins demonstrating significantly altered expression following light treatment. There was a relatively lower degree of overlap between circadian proteins and light-inducible proteins identified in this study, in contrast with our previous comparison with cycling transcripts identified by microarray gene expression profiling [11.5% (3/26) versus 19% (23/120)]. It should be noted that in the study by Deery et al., (20) only 34 circadian proteins were annotated, whereas 115 out of 871 protein “spots” in total showed significant variations over circadian time. Second, it is conceivable that the amplitude of expression at the transcript versus protein level of individual genes is disparate, resulting in some genes deemed as rhythmic in one study but not the other. Moreover, as we applied very stringent criteria for the semiquantitation based on spectral counts, a greater difference in expression must be observed in order for a protein to be considered as significantly regulated, in contrast with two-dimensional-DIGE-based quantitation, a particularly germane point in the case of higher abundance proteins. Nevertheless, our results demonstrate a clear but limited overlap between the circadian genes and light-inducible genes.

Pathway Analysis and Validation of the Light Inducible SCN Proteome

To further explore the biological relevance of the light-responsive proteome of the SCN, analysis of canonical pathways was performed by fitting the identified proteins into possible pathways using the Ingenuity Pathways Analysis (IPA). Based on the p value cutoff of 0.05, the identified proteins in this study distribute widely into 155 different pathways with at least one light-inducible protein in each (Fig. 4A and supplemental Fig. S4A). On average, ∼28 proteins per pathway can be confidently identified by the stringent quantification criteria, ∼15% of which are light-inducible. This result demonstrates that light-inducible proteins represent a relatively large portion of the SCN proteome as identified by the AutoProteome system from the SCN of a single mouse. Moreover, photic stimulation potentially influences a broad range of molecular pathways within cells of the SCN. Four canonical pathways with reported relevance to circadian rhythms and light-evoked clock entrainment were selected for further analysis: (1) the Ca²⁺/cAMP response element binding protein (CREB) transcriptional pathway; (2) the molecular clockwork; (3) the extracellular signal-regulated kinase (ERK)/mitogen-activated protein kinase (MAPK) signaling pathway; and (4) the ubiquitin-proteasome degradation pathway (Fig. 4B and supplemental Figs. S4B and C). Because of their high connection to each other as well as their direct relevance to circadian rhythms, the CREB transcriptional pathway and the molecular clockwork were combined together for the analysis. The identified proteins with significant, light-induced changes (p value < 0.05) associated with the four canonical pathways are listed in Table I.

Fig. 4.

Canonical pathway analysis and validation of identified proteins with significant alterations in expression level in the SCN in response to light. A, The summary of centralized canonical pathways with more than five light-inducible proteins (also see supplemental Fig. S4A for canonical pathways with more than one light-inducible proteins); B, Schematic representation of the CREB transcriptional pathway and molecular clockwork. Identified proteins in this study are in blue color; identified proteins with significant up-regulation and down-regulation are in red and green color, respectively; C, immunofluorescent analysis of AVP peptide expression in the SCN of light-pulsed mice and untreated controls. Low magnification images (left): scale bar = 50 μm. High magnification images (right): scale bar = 5 μm. GFAP (Glial fibrillary acidic protein), glial marker; DRAQ5, nuclear marker. Arrows indicate cells in which AVP peptide is up-regulated in response to light, and arrowheads indicate cells expressing basal levels of AVP peptide in dark controls; The mean ± S.E. relative abundance of AVP in untreated versus light-pulsed SCN samples is presented at the bottom-left of panel C; D, Western blot analysis of CSNK1D expression in SCN extract from LP- and DD-treated mice. Actin expression served as the loading control. Values presented below the blot represent the mean relative abundance of CSNK1D, normalized to actin expression. n = 6 mice per treatment condition, grouped in pools of three.

Table I. Identified proteins with significant changes in canonical pathway analysis. Canonical pathways are: CREB/molecular clockwork pathway, ERK/MAPK signaling pathway and protein ubiquitination pathway.

Canonical pathway name	IPI no.	Gene symbol	Sequence name	N-SC±S.D. (LP/DD)a	Expression/p valueb
AVP	IPI00118471.1	AVP	Vasopressin-neurophysin 2-copeptin	3.35 ± 2.71/0	↑/6.2E-3
CAMKII	IPI00124695.1	CAMK2G	Isoform 1 of Calcium/calmodulin-dependent protein kinase type II gamma chain	2.95 ± 0.75/0	↑/1.13E-2
CK1δ	IPI00138790.2	CSNK1D	Casein kinase 1, delta, isoform CRA_b	2.41 ± 1.74/0	↑/3.86E-2
Gα/Gαi/o	IPI00230192.5	GNAO1	Isoform Alpha-1 of Guanine nucleotide-binding protein G(o) subunit alpha	7.23 ± 8.05/0	↑/7.41E-05
Gγ	IPI00230194.5	GNG2	Guanine nucleotide-binding protein G(I)/G(S)/G(O) subunit gamma-2	3.38 ± 1.83/0	↑/1.13E-2
iGLUR	IPI00136967.2	GRIA2	Isoform 1 of Glutamate receptor 2	3.98 ± 2.39/0	↑/3.41E-3
mGLUR	IPI00136716.1	GRM3	Metabotropic glutamate receptor 3	2.53 ± 1.72/0	↑/2.08E-2
Integrin	IPI00132474.3	ITGB1	Integrin beta-1	2.87 ± 1.76/0	↑/2.-8E-2
PP1/PP2A	IPI00224697.1	PPP2R5E	Serine/threonine-protein phosphatase 2A 56 kDa regulatory subunit epsilon isoform	6.88 ± 0.97/0	↑/2.16E-05
PP1/PP2A	IPI00851027.1	PPP1R11	Protein phosphatase 1, regulatory (inhibitor) subunit 11	3.41 ± 1.93/0	↑/1.13E-2
E3 ring	IPI00124047.1	CUL1	Cullin-1	2.32 ± 1/0	↑/3.86E-2
E3 ring	IPI00131224.1	TCEB2	Transcription elongation factor B polypeptide 2	3.23 ± 0.99/0	↑/6.2E-3
DUB	IPI00463367.2	USP7	Isoform 2 of Ubiquitin carboxyl-terminal hydrolase 7	2.09 ± 0.62/0	↑/3.86E-2
DUB	IPI00798468.1	USP9X	Ubiquitin specific protease 9, X chromosome	11.16 ± 1.7/0	↑/4.42E-08
PSMD	IPI00457661.5	PSMD5	26S proteasome non-ATPase regulatory subunit 5	13.19 ± 6.62/5.48 ± 1.64	↑/1.6E-2
PSMD	IPI00113287.1	PSMD9	Putative uncharacterized protein	2.41 ± 1.74/0	↑/3.86E-2
RAS	IPI00113248.1	KRAS	Isoform 2A of GTPase KRas	0/1.67 ± 0.42	↓/3.86E-2
RAS	IPI00323822.1	RRAS2	Ras-related protein R-Ras2	0/2.16 ± 1.17	↓/1.61E-2
14-3-3	IPI00227392.5	YWHAH	14-3-3 protein eta	41.83 ± 12.22/53.73 ± 4.56	↓/1.31E-2
PP1/PP2A	IPI00123862.1	PPP1CC	Isoform Gamma-1 of Serine/threonine-protein phosphatase PP1-gamma catalytic subunit	2.2 ± 1.65/6.59 ± 1.98	↓/1.98E-3
E3/E3 HECT	IPI00399840.2	UBE3A	Ubiquitin protein ligase E3A	0/2.45 ± 1.09	↓/3.09E-3
PSME1	IPI00124223.3	PSME1	Proteasome activator complex subunit 1	0/3.35 ± 2.24	↓/5.94E-4

^a Normalized spectral counts (N-SC) with standard deviation (S.D.).

^b Identified proteins with significant up-regulation are labeled with up arrow; identified proteins with significant down-regulation are labeled with down arrow.

Of particular interest, 22 identified proteins were discovered to be connected with the CREB transcriptional pathway and/or the molecular clockwork (Fig. 4B). The latter, which governs circadian oscillations in nearly all cells, is comprised of interlocking positive and negative transcription/translation feedback loops that drive rhythmic expression of critical core clock components. The transcription factors CLOCK and BMAL1 bind to E-box promoter elements and activate the Period and Cryptochrome genes, whose protein products subsequently heterodimerize and translocate to the nucleus, where they repress CLOCK-BMAL1 activity and shut down their own transcription. Other rhythmic, or clock-controlled genes (ccg), are also regulated by CLOCK/BMAL1-mediated transcription. Photic regulation of this molecular loop requires, in part, on CREB, which mediates light-induced transcriptional activation of the Period genes, subsequent to its own activation by the light-triggered ERK/MAPK signaling cascade (13, 37, 53, 54). Within these pathways, seven proteins are significantly up-regulated and two proteins are significantly down-regulated following light exposure. One of the proteins is Vasopressin-neurophysin 2-copeptin, the precursor of the neuropeptide arginine vasopressin (AVP), which is expressed by SCN neurons and is both clocked-controlled (i.e. rhythmic) and regulated by the CREB transcriptional pathway (7, 21, 55). Consistent with previous observations at the mRNA level (56), our proteomics data demonstrate that nocturnal light exposure induces the expression of AVP in the SCN. To further confirm this light-induced up-regulation, we performed high-resolution confocal imaging of AVP expression in the SCN of LP- and DD-treated mice by indirect immunofluorescence (Fig. 4C). Lack of colocalization with the glial cell-specific marker glial fibrillary acidic protein (GFAP) confirmed the neuronal distribution of AVP within the SCN. AVP-positive cells were restricted to the dorsomedial aspect of the SCN (considered to be the “rhythmic ” compartment, in contrast with the retino-recipient ventrolateral aspect, of the SCN), and light stimulation markedly augmented AVP expression (Fig. 4C).

Another protein of interest from our canonical pathway analysis of the proteomics data set is Casein kinase 1 delta (CSNK1D or CK1δ). Its homolog, casein kinase 1 epsilon (CK1ε), is firmly established as a core clock component, phosphorylating the PER2 protein at a critical residue and targeting it for degradation; a mutation in CK1ε underlies the now-famous tau mutation in hamsters, which results in a dramatic shortening of the circadian period (57). However, the role of CK1δ in light-induced clock entrainment remains to be addressed. In our proteomics study, CK1δ was identified as significantly up-regulated following photic stimulation. This finding was subsequently confirmed by Western blotting (Fig. 4D), and is consistent with a previous study that noted a light-induced increase in CK1δ expression at the transcript level in the SCN following nocturnal light exposure (58).

Notably, the ubiquitin-proteasome pathway is highly enriched for light-inducible proteins (eight in total), suggesting a key role for this pathway in the regulation of photic entrainment of the circadian clock (9, 48). Interestingly, two deubiquitinating enzymes (i.e. USP7 and USP9X) were identified by our MS approach as being significantly up-regulated in the SCN following photic stimulation, whereas one ubiquitin protein ligase (i.e. UBE3A) was significantly down-regulated (as shown in Table I). As shown in Fig. 5, the light-regulated expression of USP9X and UBE3A was subsequently validated by indirect immunofluorescent analysis. A recent study reported that USP9X-mediated deubiquitination of SMAD4 is essential for TGFβ signaling (59), which in turn has been implicated in the resetting of circadian clocks in peripheral tissues (60). Furthermore, components of the TGFβ signaling pathway, including TGFβ itself and SMAD3, are expressed in the murine SCN, leaving open the possibility that USP9X regulates TGFβ signaling within the master pacemaker (61). In another study, genetic disruption of dube3a, the Drosophila homolog of UBE3A, reduced or abolished the rest and activity rhythms in flies, suggesting that expression of core proteins of the molecular clock may be affected (62). We argue that a potential interplay between deubiquitinating enzymes (e.g. USP7 and USP9X) and ubiquitin ligases (e.g. UBE3A) is essential for regulating changes in protein expression and function following photic stimulation, and that these changes, in turn, may play a role in the resetting response.

Fig. 5.

Light regulates the expression of the deubiquitinase USP9X and ubiquitin protein ligase UBE3A within the murine SCN. Immunofluorescent analysis of (A) USP9X and (B) UBE3A expression in the SCN of light-pulsed mice and untreated controls. Low magnification images (left): scale bar = 100 μm. High magnification images (right): scale bar = 50 μm. DRAQ5, nuclear marker. The mean ± S.E. relative abundance of the protein examined in untreated versus light-pulsed SCN samples is presented below the micrographs.

To further explore potential connections of the identified light-inducible proteome and their up-stream and down-stream interactions, network analysis based on the known interactions of the identified proteins was performed. This approach can highlight clusters of interacting proteins that are affected by photic stimulation. Interestingly, of the 387 proteins affected by light treatment, 352 (91%) are grouped into 19 protein interaction subnetworks with more than 13 light-inducible proteins per network (Fig. 6A, supplemental Fig. S5 and supplemental Table S6). Furthermore, 18 interaction networks were associated with at least one common protein as noted in Fig. 6A. These results demonstrate the significantly high biological correlation of the identified light-inducible proteins, predicting their possible functional correlation in response to light. For example, networks 3, 5, 7, 12, and 13 are associated with each other by at least two common proteins as highlighted in Fig. 6A. This connected network complex is highly centralized at three core nodes: huntingtin (HTT), tumor necrosis factor (TNF), and Glucose Transporter-4 (SLC2A4) (supplemental Fig. S6). The roles of these HTT-, TNF-, and SLC2A4-centered networks in circadian clock processes are currently unknown but warrant future investigation.

Fig. 6.

Network analysis and validation of the identified SCN proteins with significant changes after light treatment. A, The association of the 19 most significant networks with more than one identified protein. a, CLIC4; b, BCLAF1; c, HSPD1; d, HTRA2; e, HBD; f, VCP; g, HTRA2; h, EPB41L3; i, HTRA2; j, MYL6; k, ADRBK1; l, AIMP1; m, FLNA; n, ST13; o, TUBB3; p, TUBA4A; q, USB7; r, TUBA4A; s, TUBA4A; t, IKBKAP; u, HNF4A; v, TUBB4; w, UBR2; x, SEC11C; y, NUCB1; z, ENO2; aa, TUBB6; bb, TMEM109; cc, G3BP1; B, The first protein interaction network with 28 light-inducible proteins expressed in the SCN. The identified proteins with significant up-regulation and down-regulation are in red and green color, respectively. Detailed legends are displayed in supplemental Fig. S5; C, Immunofluorescent analysis of RASGRF1 expression in the SCN of light-pulsed mice and untreated controls. Low magnification images (left): scale bar = 50 μm. High magnification images (right): scale bar = 5 μm. Arrows indicate cells in which RASGRF1 is up-regulated in response to light, and arrowheads indicate cells expressing basal levels of RASGRF1 in dark controls; The mean ± S.E. relative abundance of RASGRF1 in untreated versus light-pulsed SCN samples is presented at the bottom-left of panel C; D, Western blot analysis of RASGRF1 expression in SCN extracts from LP- and DD-treated mice. Actin expression served as the loading control. Values presented below the blot represent the mean relative abundance of RASGRF1, normalized to actin expression. n = 6 mice per treatment condition, grouped in pools of three.

From the first network, Ras-specific guanine nucleotide-releasing factor 1 (RASGRF1), which was uniquely identified in LP samples with a normalized spectral count of 2.21, was selected for further validation (Fig. 6B). In network 1, RASGRF1 is functionally associated with the transcription factor nuclear factor kappaB (NF-kB), which has been shown to be expressed by glial cells within the SCN and modulates light-induced phase shifts of behavioral rhythms, and which may participate in cytokine-mediated modulation of the circadian system (63, 64). Biochemically, RASGRF1 is a guanine nucleotide exchange factor for specific Ras or Rho GTPases and itself is activated by various glutamate and G protein-coupled receptors, both of which are important for light-induced SCN clock entrainment (65). As shown in Fig. 6C, there is significant up-regulation of RASGRF1 in SCN neurons following photic stimulation. Western blotting revealed a ∼twofold increase in RASGRF1 expression in the SCN in response to light (Fig. 6D). Although RASGRF1 has not been previously implicated in the regulation of circadian rhythms, based on its biochemical profile, one may speculate that it plays a role in the intracellular signaling events that couple light-induced activation of cell surface receptors to changes in Ras/MAPK-dependent gene transcription or to possible Rho-mediated structural remodeling of SCN neurons. Moreover, genetic ablation of RASGRF1 in mice impairs hippocampal-dependent learning and memory consolidation (66), raising the possibility that RASGRF1 may serve a similar role in the SCN in controlling neuronal plasticity in response to light.

In summary, the AutoProteome system represents the first fully automated and integrated proteomics platform that incorporates all requisite sample processing steps and online two-dimensional LC-MS/MS analysis. It should be noted that this system is essentially based on a standard LC-MS system, which can easily be adapted and applied for different large-scale proteomics studies (i.e. phosphoproteomics) where the starting material is in limiting quantities. In addition, we report the first light-responsive proteome of the master circadian pacemaker in mammals, the SCN, using tissue harvested from a single mouse for each experiment. Broad distribution into different key canonical pathways and a tight connection of the entire light-responsive proteome demonstrate the potential biological significance of our findings to the regulation of mammalian circadian rhythms. Lastly, from our proteomics results, we confirmed the light-inducible nature of two proteins that had previously been implicated in clock timing processes within the SCN, as well as three other proteins which are as yet uncharacterized with respect to their role in circadian rhythms.