A day in the life of a gene.

Ensembl_ID

Ensembl gene identifier.

Coordinates

Gene location in the mouse mm10 genome.

Gene_ID

Gene symbol used in the study.

Gene Name

Descriptive gene name.

FPKM

Fragments per kilobase per million mapped fragments: an expression measure adjusted for transcript length and sequencing depth. Larger values indicate greater expression in the SCN. Generated using cuffQuant and normalised with cuffNorm.

Expression level (counts)

Mean number of reads mapped to the gene over 24 hours. Counts depend on gene length; FPKM is more suitable when comparing expression between different genes.

SCN enriched

Whether expression meets the study’s criteria for significant enrichment in the SCN relative to brain tissue; see Methodology.

Enrichment

Fold-change difference between SCN and brain expression. A value of 10 indicates ten-fold higher expression in the SCN.

Fluctuating

Whether the gene is defined as significantly fluctuating over time; see Methodology.

JTK.padj

Adjusted p-value for sinusoidal cycling from JTK_CYCLE. Values below 0.05 indicate significant cycling under the study’s criterion.

DESeq2.padj

Adjusted p-value for expression change over time. Values below 0.05 indicate significant change.

WGCNA module

Co-expression module assigned using WGCNA. The twin-peaking genes form the lightsteelblue1 module.

Peak phase

Estimated time of peak expression, in Zeitgeber hours. This statistic is most applicable to sinusoidally cycling genes.

Peak phase (binned)

Peak-expression time grouped into six bins: ZT2, ZT6, ZT10, ZT14, ZT18 and ZT22.

Dynamic range

Fold-change difference between peak and trough expression over time.

SCN Panda et al.

Cycling statistic from the earlier SCN microarray study (Panda et al., 2002). Values below 0.1 suggest significant sinusoidal cycling in that study.

Tissue Cyc No

Number of other tissues (maximum 12) showing sinusoidal cycling with adjusted JTK p-value below 0.05 in Zhang et al. (2014).

Expression quantification

Aligned reads were quantified using HTSeq (Anders et al., 2015) or cuffNorm. Both methods used a GTF combining the Ensembl annotation and the identified lincRNAs. FPKM values were generated with cuffQuant and normalised with cuffNorm; count tables were generated with HTSeq and normalised using DESeq2 (Love et al., 2014).

Identification of SCN-enriched genes

Additional RNA-seq datasets from six independent studies were used: GSE54124, GSE30352, GSE54652, GSE41637, GSE41338 and GSE36874. Genes were required to show significantly greater expression in brain than liver, and in both this study’s SCN dataset and the published SCN dataset (Azzi, 2014) relative to brain, using ANOVA in R (q < 0.05). FPKM values were averaged for each study and quantile normalised. Fold-change scores compared mean expression between tissues. SCN enrichment required a three-fold enrichment over brain samples.

Identification of fluctuating genes

JTK_CYCLE (Hughes, 2010) and DESeq2 identified genes whose expression changed over time. DESeq2 likelihood-ratio tests assessed temporal change (q < 0.05), and JTK_CYCLE assessed 24-hour periodic waveforms (q < 0.05). Signed WGCNA (Langfelder, 2008) identified co-expression modules, merged when their similarity exceeded 0.3 by dendrogram height. Of the resulting 23 modules, lightsteelblue1 was investigated further as the robust non-sinusoidal, twin-peaking module. Its genes were defined as fluctuating when DESeq2 showed significant temporal change (q < 0.05).

Identification of lincRNAs

CGAT NGS pipelines rnaseqtranscripts.py and rnaseqlncrna.py (Sims, 2014) were used. The first identified transfrags with Cufflinks and retained those present in at least two samples. The second removed transfrags overlapping protein-coding exons. Coding potential was assessed with the coding potential calculator (CPC; Kong, 2007), removing candidates annotated as coding (CP score > 1). Intergenic candidates more than 2 kb from a protein-coding gene, with multiple exons and expression in more than six biological replicates, were retained for investigation.

Identification of fluctuating genes

The Gene table Methodology section describes the temporal-expression analyses.

Nearest PC L / R

Closest protein-coding gene on the lincRNA’s left or right. L means lower genomic coordinates; R means higher coordinates.

Distance PC L / R

Distance from the lincRNA to the neighbouring protein-coding gene, in base pairs.

FPKM

Fragments per kilobase per million mapped fragments: an expression measure adjusted for transcript length and sequencing depth. Larger values indicate greater expression in the SCN. Generated using cuffQuant and normalised with cuffNorm.

Expression level (counts)

Mean number of reads mapped to the lincRNA over 24 hours. Counts depend on length; use FPKM for comparisons between different transcripts.

Fluctuating

Whether the lincRNA is defined as significantly fluctuating over time.

JTK.padj

Adjusted p-value for sinusoidal cycling; values below 0.05 indicate significant cycling.

GLM.padj

Adjusted p-value for temporal change from the DESeq2 count model; values below 0.05 indicate significant change. This is the source table’s GLM.padj column, called DESeq2.padj in the archive.

WGCNA module

Co-expression module. The twin-peaking genes form the lightsteelblue1 module.

Peak phase

Time of peak lincRNA expression, in Zeitgeber hours; most applicable to sinusoidally cycling genes.

Dynamic range

Fold-change difference between peak and trough lincRNA expression over time.

Correlation PC L / R

Pearson expression correlation between the lincRNA and its nearest protein-coding neighbour on either side.

lincRNA phastCons

Conservation score for the lincRNA: larger values indicate stronger conservation. Values near 1 are more conserved; values near 0 are less conserved.

TE phastCons

Conservation of nearby transposable-element sequence, used as a neutral-evolution comparison.

z-score cons - TE

Conservation relative to nearby transposable elements within 50 kb. As described in the original site, a value of 3 means three-fold conservation relative to that background.

Flanking phastCons

Conservation of flanking sequence, used as another neutral-evolution comparison.

z-score cons - flanking

Conservation relative to a 500 bp flanking region 2 kb from the lincRNA locus.

Gene name

Previously assigned name, if the lincRNA was annotated.

Annotated

Whether it matches the existing annotations used in the study.

Belgard

Previously identified in mouse cortex by Belgard et al. (2011).

Ramos

Previously identified in neural stem cells by Ramos et al. (2013).

CAGE permissive

Cap Analysis of Gene Expression evidence for a transcription start site at the locus.

Retro pseudogene overlap

Whether the lincRNA overlaps a retroposed pseudogene.

TE overlap

Whether the lincRNA overlaps a transposable element.

Human linc

Candidate human lincRNA identifier in the corresponding genomic neighbourhood.

Name / Coordinates

Study-assigned NONCO identifier and genomic position in the mouse mm10 assembly.

Six time-point means from the archived study workbook, labelled RPKM. These are separate from the published FPKM and count summaries in the table. Replicate measurements and time-point-specific standard deviations were not recovered, so no error bars are shown.

Archived RPKM time-point means · light ZT0–12 · dark ZT12–24 · no SD available

Workbook records were matched to all 3,187 published lincRNAs by identifier and exact coordinates; cycling p-values, phases and modules also agree. This establishes the identity of the records, not equivalence to the final count normalisation.

Identification of novel exons

De novo transcript assembly used Cufflinks v2.0.2 (Trapnell et al., 2010), identifying 6,610 novel exons that did not overlap Ensembl or UCSC annotations but lay within 10 kb of an annotated feature. Retained exons had less than 25% reciprocal overlap with retroposed pseudogenes (ucscRetroAli6), less than 50% reciprocal overlap with transposable elements, more than 20 spliced reads in the novel exon, more than 10 spliced reads per 100 bp, more than 20 reads splicing into a known transcript, and length below 3,000 bp. These filters yielded the robust set of 1,013 novel exons.

SCN RNA sample preparation

Male C3H/HeH mice aged 10–12 weeks were group housed in light-tight chambers with LED lighting at 150 lux at the cage floor. Following seven days of acclimatisation, mice were singly housed for seven days before tissue collection. Brains were collected at one of six Zeitgeber times and immediately frozen on dry ice in OCT mounting medium.

Eighteen frozen coronal sections, 15 μm thick, spanning the rostral-to-caudal SCN were mounted on polyethylene naphthalate (PEN) slides. Laser-capture microdissection of dehydrated, Nissl-stained sections used the PALM system, as previously described (Dulneva et al., 2015).

Tissue from five animals was pooled for each biological replicate. Total RNA was purified with the RNeasy micro kit (Qiagen); RNA Picochip (Agilent) assessment gave RIN values above 8 and an average yield of approximately 10 ng per replicate. Approximately 1 ng was amplified using the SMARTer protocol. Libraries were sequenced as 100 bp paired-end reads on HiSeq (Illumina; Wellcome Trust Centre for Human Genetics Sequencing Core), yielding approximately 35 million read pairs per technical replicate and approximately 100 million per biological replicate.

Additional methodology appears in the Gene table, SCN lincRNA and Novel exons sections.