henhaobio
Sign in

bioRxivGenomicsPreprint

Circadian transcriptome in cultured cells depends on synchronisation method and differs from in vivo condition

Michal DudekiD, Cátia F. GonçalvesiD, Judith Alison HoylandiD, Qing‐Jun MengiD

PDFDOICodeData
In-depth full-text digest available · Jump to digest ↓

Abstract

In vitro synchronisation is widely used to study circadian clocks in cells, but whether cultured cells recapitulate tissue-level rhythmic outputs remains unclear. Articular cartilage provides a useful model to address this because chondrocytes are the only resident cell type. Here, we compared circadian time series transcriptomes between primary mouse chondrocytes synchronised by heat shock, dexamethasone, or osmotic stress and in vivo cartilage tissue. All three stimuli robustly synchronised core clock gene rhythms but produced distinct circadian phases and markedly different rhythmic transcriptomes, depending on the synchronizer. Heat shock, dexamethasone, and osmotic stress yielded 5255, 2008, and 879 transcripts classified as rhythmic, respectively, in primary chondrocytes, yet only 64 genes were shared across the three in vitro datasets, and only 15 were shared when in vivo cartilage transcriptome was included. Pairwise comparisons between synchronizers revealed some statistically enriched overlaps, but only marginally above chance, and shared genes showed limited conservation of circadian phase. Functional enrichment analysis also revealed stimulus-dependent rhythmic programmes with modest pathway-level overlap. These findings indicate that circadian output in cultured cells is shaped by the synchronising cue and cellular microenvironment. We also present BodyClocks.org, an interactive resource implementing this comparative framework across a curated collection of circadian transcriptomic datasets.

In-depth summary

Claude Opus 5.5 · xhigh effort · 7 Nov 2026 · from the full text of version 1 (23 Jul 2026) · CC BY 4.0

Heat shock, dexamethasone and osmotic stress all entrain the clock of primary mouse chondrocytes but yield distinct rhythmic transcriptomes, overlapping only marginally with each other and in vivo cartilage.

Synchronised cell cultures are routinely used to infer a cell type's clock-controlled genes, yet whether they reproduce the rhythms of intact tissue is unresolved; cartilage, whose only resident cells are chondrocytes, allows a direct test. The study compares three synchronisers against in vivo tissue and, unlike gene-centric resources such as CircaDB and CircaKB, releases a tool for pathway-, network- and dataset-level comparison of circadian transcriptomes.

Key results

  1. ShownAll three cues re-synchronised PER2::LUC rhythms in chondrocytes, yet the size of the rhythmic transcriptome depended strongly on the cue and only a small core, dominated by clock genes, was rhythmic in every dataset.

    Rhythmic transcripts (RAIN, BH.Q < 0.05): heat shock 5255, dexamethasone 2008, osmotic stress 879. 64 genes shared by the three in vitro sets; 15 shared once in vivo cartilage was added, including Bmal1, Npas2, Cry1, Nr1d1, Per1, Per2, Per3, DBP and Tef, whose temporal patterns differed between synchronisers. · Abstract; RNA sequencing data processing and circadian…; Different synchronising cues generate robust but…

  2. ShownPairwise overlaps of rhythmic gene lists exceeded chance only slightly, and no comparison involving osmotic stress reached significance.

    Hypergeometric tests: overlap significant (BH-adjusted p ≤ 0.05) for heat shock (HS) vs dexamethasone (Dex), HS vs cartilage and Dex vs cartilage, but only 1.08-fold (HS vs Dex) to 1.2-fold (Dex vs cartilage) above chance; Jaccard 0.145 and 0.093 for these two pairs. At BH.Q < 0.01, only Dex vs HS (1.27-fold) and HS vs cartilage (1.63-fold) were significant. · Rhythmic genes show low overlap between…; Comparative overlap analysis between datasets

  3. ShownWith phase expressed relative to Bmal1, shared rhythmic genes in pairs among dexamethasone, heat shock and cartilage were more phase-concordant than chance, whereas all pairs involving osmotic stress were not; overall concordance was modest.

    Concordance defined as clock-relative phase (CRP) distance ≤ 4 h, tested against 1,000 CRP permutations. Of the 64 genes rhythmic in all three in vitro sets, 23 were concordant (35.9% vs permutation null 19.4%, perm_p ≤ 0.001). Absolute phase distributions and core clock gene phases differed among the four datasets. · Shared rhythmic genes show limited conservation…; Comparative overlap analysis between datasets

  4. ShownOverlap at the pathway level was modest, and whether it was significant depended on the choice of enrichment background.

    With pair-specific tested-gene backgrounds, 6 of 18 dataset-pair by annotation-category comparisons had no enriched term in at least one dataset, 15 had at most one, and no overlap survived correction. With all STRING-mapped mouse genes as background, overlap of enriched KEGG, Reactome and WikiPathways terms was significant in 11 of 18 comparisons (permutation test, BH). · Pathway-level overlap remains modest and…; Comparative overlap analysis between datasets

  5. SuggestedAgainst public mouse datasets in BodyClocks, cultured chondrocyte rhythmic gene sets tended to overlap in vivo tissues less than in vivo tissues overlap one another, but the difference did not survive multiple-testing correction.

    Mean fold enrichment 1.083 for chondrocyte-in vivo pairs against 1.391 for in vivo-in vivo pairs (figure legend: n = 66 and n = 231; chondrocyte-chondrocyte mean 1.06, n = 3); exact permutation p = 0.0091 nominal, BH-adjusted p = 0.0548. Across 25 mouse datasets, the chondrocyte sets clustered in a low-enrichment block alongside dexamethasone-synchronised podocytes. · Rhythmic genes show low overlap between…; BodyClocks enables interactive exploration and…; Figure legends

Methods, data and reproducibility

PER2::LUC primary mouse chondrocytes, dexamethasone-pre-synchronised then reset after 112 h by heat shock or a second dexamethasone pulse, were sampled every 4 h for 48 h (three time courses) for RNA-seq. Rhythmic genes (RAIN, BH.Q < 0.05) were compared with osmotic-stress and in vivo cartilage datasets for overlap, Bmal1-relative phase and pathway enrichment, a workflow packaged in the BodyClocks app.

  • Primary articular chondrocyte circadian RNA-seq time series after heat shock or dexamethasone resetting (this study; 5-day-old PER2::LUC mice; litter-distinct replicates) · mouse · n = 3 independent 48 h time courses per condition, sampled every 4 h · E-MTAB-17163
  • Osmotic stress-synchronised primary chondrocyte circadian dataset (earlier study) · mouse · E-MTAB-11040
  • In vivo articular cartilage circadian transcriptome (previously published) · mouse (inferred: BodyClocks compares only same-species datasets) · E-MTAB-3428
  • Curated public circadian transcriptomic datasets processed for BodyClocks (25 mouse datasets used in the cross-dataset fold-enrichment heat map) · mouse and baboon · n = over 90 datasets
  • Zenodo archives of analysis code/data and of the BodyClocks app source · 10.5281/zenodo.21379611; 10.5281/zenodo.21379114
Methods & tools
PER2::LUC bioluminescence (LumiCycle, Actimetrics), TruSeq Stranded mRNA, HiSeq 4000 (Illumina), Trimmomatic, STAR, DESeq2, ComBat-seq (sva), RAIN, STRING / STRINGdb, biomaRt, ComplexHeatmap, R Shiny, visNetwork
Software & environment
bcl2fastq v2.20.0.422, FastQC v0.11.3, MultiQC v1.8, Trimmomatic v0.39, STAR v2.7.3a (GRCm38/mm10, GENCODE M25), DESeq2 v1.50.2, STRING v12.0 (confidence ≥ 700), R Shiny ≥ 1.11.1; Docker container provided for local BodyClocks deployment.
Reproducibility gaps
The osmotic-stress synchronisation protocol is not described here (deferred to the study behind E-MTAB-11040); the E-MTAB-17163 link is printed with an access key parameter.
Limitations (5)
  • The osmotic-stress and in vivo cartilage datasets came from the group's earlier studies and may differ from the new heat shock and dexamethasone data despite shared cell preparation and sampling protocols; all data were reprocessed through one pipeline to reduce this. · stated by the authors
  • With pair-specific tested-gene backgrounds, enriched-term lists were too sparse to give power for pathway comparison; conclusions shift with the enrichment universe, so results for both backgrounds are reported. · stated by the authors
  • The lower fold enrichment for chondrocyte-in vivo than in vivo-in vivo comparisons was nominally significant (p = 0.0091) but not after BH correction (adjusted p = 0.0548). · stated by the authors
  • Rhythmicity was called with RAIN in longitudinal mode for chondrocytes but independent mode for cartilage, and the in vivo reference is a single cartilage dataset, so part of the in vitro vs in vivo difference could reflect analysis or dataset choice. · noted in this summary
  • The proposed drivers of stimulus-specific output (glucocorticoid receptor targets, heat shock proteins, mTOR/NFAT5/MAP kinase signalling) are discussed but not tested experimentally. · noted in this summary

Read this if you use synchronised cell cultures to infer a tissue's clock-controlled genes, or want a web tool for comparing rhythmic gene sets, phases and pathways across circadian transcriptome datasets.

Citations
0

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.

Related papers

Similarity Map
SelectedSame fieldCross-field
THISPritt '24Vago '21Goncalves '24Nakamura '21Rzechorzek '24Gould '18
66% semantic matchGenomics
Jacob Pritt, Julie D. Counts, Milton Campbell+2 more·bioRxiv·Jan 20, 2024