The proteomic landscape of the suprachiasmatic nucleus clock reveals large-scale coordination of key biological processes

Abstract

The suprachiasmatic nucleus (SCN) acts as the central clock to coordinate circadian oscillations in mammalian behavior, physiology and gene expression. Despite our knowledge of the circadian transcriptome of the SCN, how it impacts genome-wide protein expression is not well understood. Here, we interrogated the murine SCN proteome across the circadian cycle using SILAC-based quantitative mass spectrometry. Of the 2112 proteins that were accurately quantified, 20% (421 proteins) displayed a time-of-day-dependent expression profile. Within this time-of-day proteome, 11% (48 proteins) were further defined as circadian based on a sinusoidal expression pattern with a ∼24 h period. Nine circadianly expressed proteins exhibited 24 h rhythms at the transcript level, with an average time lag that exceeded 8 h. A substantial proportion of the time-of-day proteome exhibited abrupt fluctuations at the anticipated light-to-dark and dark-to-light transitions, and was enriched for proteins involved in several key biological pathways, most notably, mitochondrial oxidative phosphorylation. Additionally, predicted targets of miR-133ab were enriched in specific hierarchical clusters and were inversely correlated with miR133ab expression in the SCN. These insights into the proteomic landscape of the SCN will facilitate a more integrative understanding of cellular control within the SCN clock.

Figure 1. Global proteomic analysis of the murine SCN.

(A) Schematic overview of the centrifugal proteomic reactor (CPR) coupled with SILAC-based quantification of the murine SCN proteome. Protein lysates (30 µg) extracted from the SCN of individual mice (n = 4 per CT; 6 CT in total) were mixed with equal quantities of protein lysates from heavy SILAC-labeled Neuro2A cells. The mixtures were processed by the CPR coupled with HPLC-ESI-MS/MS. A total of 3275 unique proteins were identified, of which 421 were significantly altered (time-of-day-dependent) in terms of protein expression levels during a 24-h cycle. (B–D) Biological replicates within a CT (B,C) showed a higher degree of correlation than samples harvested at different CTs (D). Scatter plots were plotted by logarithmized (Log₂) normalized protein ratios (L/H) and the correlation coefficient (Pearson r) was calculated. (E,F) GO enrichment analysis by DAVID based on (E) total cellular component and (F) total biological process. The significantly altered (blue), circadian (red), and total SCN (black) proteomes were subject to enrichment analysis. All listed classifications were significant compared to the whole genome. Asterisks denote classifications that were significantly enriched compared to the SCN proteome (total). *p<0.05, **p<0.01, ***p<0.001 (Fisher's exact test).

Figure 2. Robustness and validation of our SILAC-based SCN proteome.

(A–E) Raw MS results of the temporal profiles of 12 time-of-day-dependent proteins including proteins that exhibit (A) 8 h, (B) 12 h, and (C) 24 h rhythms. (D,E) Time-of-day-dependent proteins, including those that are encoded by 24 h rhythmic transcripts (D), are also shown. Each graph was plotted with quantification values (Log₂ (L/H)) in all 24 samples. The median value ± SEM of 4 biological replicate measurements for each CT (n = 4 per CT; 6 CT in total) is also shown. Time-of-day-dependent expression of SH3GL2, VAMP2, PAK1, SYT1 and SV2A were validated by Western blot (WB) analyses and presented below each MS plot. WB expression was analyzed at 6 CTs (2, 6, 10, 14, 18, 22). Actin was used as the loading control. Values below each blot represent the median relative abundance of the protein of interest, normalized to actin expression (n = 3 per CT). The asterisk (*) in panel (E) denotes the presence of a faster-migrating, non-specific band.

Figure 3. Cluster analysis of the time-of-day-dependent SCN proteome.

(A) Hierarchical clustering of the 421 proteins that exhibited statistically significant, time-of-day-dependent expression in the SCN. After z-score normalization of the median value of logarithmized intensities (Log₂) of each protein profile within Euclidean distances against those 421 time-of-day-dependent proteins, they were classified into six different expression clusters (denoted A through F). (B) Expression profile of the six hierarchical clusters, which were statistically different relative to one another. Two dominant clusters, B and E, were mirror images of one another. (C) Distribution of GO biological process terms in the six hierarchical clusters. Three GO biological processes were specifically enriched in cluster E relative to the time-of-day proteome. **p<0.01, ***p<0.001.

Figure 4. Comparative analysis of the murine SCN transcriptome and proteome.

(A–D) Distribution of the (A) time-of-day proteome, (B) circadian proteome, (C) 12-h ultradian proteome, and (D) 8-h ultradian proteome according to profile of transcript expression. Transcript profile was classified as non-rhythmic (green), 24 h rhythmic (blue), or rhythmic non-24 h (orange). The rhythmic non-24 h transcripts oscillated at periods of either 16, 20, 28 or 32 h. Transcript data were acquired from two published microarray studies (MAS4 Panda et al and gcrma Panda et. al) of the mouse SCN transcriptome from the CIRCA database. Proteins without a corresponding transcript in CIRCA database are not represented in the pie charts. (E) Genes that are circadian at the transcript and protein level: a comparison of the phases of peak expression. Pink and blue dots represent the phase (CT) of peak expression at the mRNA and protein level, respectively. In general, expression of these 24 h rhythmic proteins lagged significantly behind expression of their corresponding transcripts. Asterisks (*) denote instances where the transcript and protein overlap in their phase of peak expression.

Figure 5. Functional implications for miRNA target enrichment in specific hierarchical clusters.

(A) Distribution of the number of predicted miR-133ab murine target genes in the six hierarchical clusters. Compared to other hierarchical clusters, cluster E was significantly enriched for predicted murine targets of miR-133ab (Fisher's exact test, p<0.05). (B) Expression profiles of miR-133a and miR-133b, along with (C) those of their respective protein targets in cluster E. qRT-PCR was performed to detect and measure the relative abundance of miR-133a and miR-133b in the SCN. Information regarding predicted miRNA targets was extracted from the broadly conserved microRNA families from the TargetScanMouse database. (D–F) Western blot analysis of three predicted miR-133ab targets in cluster E, including (D) SH3GL2, (E) SYT1, and (F) SV2A, in Neuro2A cells transfected with either microRNA inhibitors against miR-133a (A-) or miR-133b (B-), microRNA mimics for miR-133a (A) or miR-133b (B), or microRNA inhibitor negative controls (C). Actin was used as the loading control in Western blot analysis. Values in each graph represent the median relative abundance of the protein examined normalized to actin expression (n = 3 per CT). The asterisk (*) denotes the presence of a faster-migrating, non-specific band.

Figure 6. Protein interaction network of cluster E proteins.

A total of 35 proteins from cluster E had direct protein-protein interactions based on IPA network analysis. Of note, proteins involved in neurotransmitter release (#) and synaptic transmission ($) were observed in this network.

Figure 7. Mitochondrial oxidative phosphorylation represents a major axis of regulation within the SCN.

(A) KEGG pathway enrichment analysis (by DAVID) of the time-of-day proteome. Pathways that are significantly enriched relative to the SCN proteome (2112 proteins) are denoted with an asterisk. *p<0.05, and **p<0.01 (Fisher's exact test). (B) Schematic representation of the 19 rhythmic proteins in our proteomic dataset that are involved in oxidative phosphorylation, based on IPA canonical pathway analysis. The asterisk (*) denotes four 24-h rhythmic proteins including NDUFA10, NDUFA2, COX4I1, and ATP5D. (C) Immunofluorescent (top) and MS (bottom) analysis of NDUFA10 expression in the SCN as a function of CT. IF images were acquired using a 20× objective. Values below each micrograph (top) represent the median relative abundance of NDUFA10 (n = 3 mice per CT). (D) Distribution of the 111 time-of-day-dependent mitochondrial proteins within the six hierarchical protein clusters.