Harmonics of Circadian Gene Transcription in Mammals

Your liver runs on a twenty-four hour clock — that much everyone knows. But buried in a high-resolution dataset, Hughes and colleagues found clusters of genes in mouse liver cycling every twelve hours and others every eight. Two beats per day, three beats per day, layered on top of the fundamental rhythm. And here's the thing that makes that finding genuinely strange: pull the liver out of the animal, put it in a dish, and those faster rhythms vanish entirely. The twenty-four hour clock keeps ticking. The harmonics disappear. That tension is the center of this paper. And to understand why it matters, you need to grasp the scale of what the circadian clock actually does. Circadian rhythms are twenty-four hour oscillations in physiology and behavior — body temperature, blood pressure, metabolism, locomotor activity. At the molecular level, a cellular clock built from transcriptional feedback loops generates these oscillations. In mammals, the central pacemaker sits in two small clusters of hypothalamic neurons called the suprachiasmatic nuclei, or SCN. But most tissues also contain their own peripheral clocks — in liver, heart, kidney, and lung — capable of sustained oscillations and regulated by signals from the SCN, including glucocorticoids, core body temperature, and cyclic AMP. Because rhythmic physiology is thought to arise from the transcriptional output of these clocks, mapping which genes cycle, where, and when is essential. The catch is that most mechanistic work on the clock had been done in cultured cell lines — NIH three T three fibroblasts and U two O S osteosarcoma cells, primarily — because they're controllable and reproducible. No one had done a rigorous, genome-wide, high-resolution comparison between cells in a dish and cells in a living animal. That's the gap Hughes and colleagues set out to close. The methodological choice that made this study work was sampling density. Instead of collecting samples every two or four hours — the standard in earlier circadian transcriptomics — the team collected mouse liver samples every hour for forty-eight hours. The same one-hour resolution was applied to synchronized NIH three T three and U two O S cells in parallel. NIH three T three cells were synchronized with forskolin, U two O S cells with dexamethasone, and all samples were run on Affymetrix microarrays. To call a transcript rhythmic, the team used two independent statistical tests — COSOPT and Fisher's G-test — and required agreement between both at a false discovery rate below zero point zero five. That combination of dense sampling and two-algorithm cross-validation is what revealed what came next. The first major result is the number. In mouse liver, both algorithms agreed on three thousand six hundred sixty-seven transcripts cycling with approximately twenty-four hour periods at that false discovery rate threshold. In NIH three T three and U two O S cells, using identical analysis, fewer than a dozen. That is not a modest difference. That is a hundred-fold gap between the transcriptional output of a clock running inside an intact organ and the same clock running in a dish. Core circadian clock genes — Per family members and Nr1d family members — do oscillate in both cell lines. Reporter assays confirmed the cells have functioning, synchronized circadian rhythms. But the downstream, clock-controlled transcriptional program — the thousands of genes that a liver clock actually regulates — is essentially absent in culture. The clock ticks, but it's not driving much. And crucially, the amplitudes of even core-clock transcripts are consistently dampened in the cellular models relative to liver. Hughes and colleagues frame this directly as a challenge to standard practice. Results from cell-based circadian studies may not capture the physiologically relevant transcriptional program of the whole animal. That's a pointed message for a field that has relied heavily on those models. Now for the headline finding. Alongside the three thousand six hundred sixty-seven transcripts cycling at twenty-four hours, the same analysis found two hundred sixty transcripts cycling at approximately twelve hours and sixty-three cycling at approximately eight hours. These are harmonics of the circadian period — the same word you'd use in music, where a note has overtones at twice and three times its base frequency. To confirm these weren't twenty-four hour genes misclassified, the team re-fit the twelve hour and eight hour gene sets while restricting permitted periods. For the twelve hour set, fitting a twelve hour model produced median p-values of zero point zero zero one. Fitting a twenty-four hour model to those same genes produced median p-values of zero point four. The eight hour set was even more decisive: the circadian fit gave median p-values of one point zero. These are genuinely faster rhythms, not statistical artifacts of the detection method. Experimental validation backed this up. An independent liver time course analyzed by quantitative PCR reproduced both the core clock rhythms and the sub-circadian period lengths. One transcript in particular — Hspa one b, encoding the HSP seventy protein — showed clear twelve hour rhythms in every tissue tested, with nearly identical phase across tissues. Twelve hour oscillations were validated in heart, kidney, and lungs — at least six tissues in total. The harmonics are not a liver-specific quirk. They're a feature of peripheral physiology more broadly. Why had nobody seen them before? Because nobody had sampled densely enough. Hughes and colleagues ran a simulation: take the NIH three T three and U two O S hourly datasets and down-sample them to four-hour resolution. At that coarser resolution, COSOPT declares thousands of transcripts rhythmic — five to ten percent of the genome — consistent with earlier studies that used similar sampling intervals. One-hour sampling changes both the p-values and q-values and produces a more conservative, rigorous picture. Finer sampling is what reveals the faster rhythms and reduces false positives at the same time. That brings us to the mechanistic question: what produces the harmonics? The contrast between intact liver and cultured cells points directly at systemic signals. In cultured NIH three T three and U two O S cells, the harmonics are absent. In ex vivo primary hepatocytes — cells freshly isolated from mouse liver but then cultured — the twenty-four hour clock persists, but twelve hour transcripts are either severely dampened or entirely absent. Some were dramatically reduced; others were gone. The autonomous clock can maintain its fundamental frequency on its own. The harmonics require the organism. To test whether feeding is a specific driver, Hughes and colleagues used a restricted feeding paradigm: food available only during an eight-hour window in the subjective day, when mice normally don't eat. Under this regimen, seven of eight tested twelve hour transcripts dramatically changed their expression patterns, and one became entirely arrhythmic. The transcripts that remained rhythmic shifted to a single peak coinciding with the imposed feeding window — the normal subjective evening peak was largely gone. Their conclusion is direct: at least one component of the twelve hour rhythms is driven by feeding. The mechanism they point toward involves feeding behavior and food metabolism, possibly acting through endoplasmic reticulum stress pathways, as a systemic driver that reinforces these twice-daily rhythms in vivo. Three lessons crystallize from this work. First, cell-based circadian models miss most of the clock's transcriptional output. The core clock is intact in culture. The biology downstream of it is largely not. Second, sampling density is not a technical nicety — it is what made an entire layer of gene regulation visible. Without one-hour resolution, the twelve hour and eight hour cycles stay hidden. And third, the circadian system is not just a cell-autonomous oscillator that runs the same everywhere. It integrates systemic signals — feeding, metabolism, whatever cues connect organ to organism — and those signals generate transcriptional rhythms that the isolated cell simply cannot produce. What physiological processes follow twice-daily transcriptional programs is an open question this paper raises but doesn't fully answer. Digestion cycles roughly twice a day. Metabolic processes tied to feeding have obvious twelve hour structure. The presence of twelve hour rhythms in heart and kidney as well as liver suggests something coordinated across the body, not just a local response to food in the gut. Hughes and colleagues give the field the tools and the framework to start asking those questions precisely — hourly sampling, two-algorithm validation, and the clear instruction to look for harmonics that previous methods would have missed entirely. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.