Numerous processes in nature possess rhythmic and repetitive oscillations within a specific time period, ranging from calcium oscillations in cells and the sleep-wake cycle to the reproductive cycle (Goldbeter, 2008). Biological rhythms include the circadian rhythm that has a period of approximately 24 h, the infradian rhythm that has a period longer than 24 h, and the ultradian rhythm that occurs multiple times within 24 h (Dibner et al., 2010; McGinnis and Young, 2016). While biological rhythms can be generated by self-sustained oscillators that are maintained in the absence of environmental cues, external signals such as photoperiod or hormonal signals can entrain the rhythm (Gachon et al., 2004).

The most characterized biological rhythms so far are circadian rhythms. The 2017 Nobel Prize in Physiology or Medicine was awarded to three researchers for the discovery of molecular mechanisms controlling the circadian rhythm. The molecular circadian clock is composed of a transcriptional-translational feedback loop of activators and repressors that can control the rhythmicity of their own expression (Takahashi, 2017). In mammals, the suprachiasmatic nucleus (SCN) in the hypothalamus has been identified as the master circadian pacemaker (Hastings et al., 2018; Takahashi, 2017). The molecular machinery of the circadian clock is present in many plant, animal, fungal, and cyanobacterial species, and is similar in principle (Edgar et al., 2012).

In contrast to the circadian rhythm, the central pacemaker and molecular mechanisms of ultradian rhythms have not yet been elucidated. One of the well-known examples of an ultradian rhythm includes the hypothalamic gonadotropin-releasing hormone (GnRH) pulse generator that governs pituitary secretion of gonadotropins, which is important for the regulation of mammalian reproduction (Herbison, 2018; McCartney and Marshall, 2014). Kisspeptin neurons in the hypothalamic arcuate nucleus (ARN) appear to be essential components that control GnRH neurons in the initiation and maintenance of development and reproduction, but the origin of the pulse generator remains elusive. Here, we briefly review the ultradian rhythm with a special reference to the calcium oscillations in the hypothalamic ARN and from a developmental perspective.

Ultradian rhythms maintain the stability of living organisms by preparing the system to respond and adapt quickly to external signals. These oscillations can be considered as homeodynamic conditions that involve constantly changing self-organization processes (Yates and Yates, 2008). Cellular processes such as energy production, biosynthetic pathways, protein assembly, stress response, and cell differentiation/division depend on the coordination of ultradian organization (Cortassa et al., 2011). Homeodynamic conditions allow living systems to coordinate their function in a spatio-temporal manner and organize the complexities of life (Lloyd et al., 2001). Loss of ultradian rhythmicity can lead to dysregulation of physiological processes. One long-studied example is related to reproduction. In hypothalamic-lesioned monkeys, intermittent delivery of GnRH established pulsatile gonadotropin secretion, while constant infusion failed to restore gonadotropin secretion (Belchetz et al., 1978). Discovery of the mode of GnRH secretion has been utilized in therapeutics for infertility such as hypogonadotropic hypogonadism.

Another prominent example is the blood level of glucocorticoid (GC). Often, ultradian rhythms are related to circadian rhythms. GC is an adrenal steroid secreted from the adrenal gland, which has a peripheral clock and generates daily oscillations of GC (Son et al., 2008). Interestingly, in the pattern of GC secretion, an hourly ultradian pulsatility exists together with a circadian rhythm to maintain stress responsivity and to prevent the downregulation of GC signaling systems (Lightman and Conway-Campbell, 2010). Nonpulsatile GC rhythm changes the neural processing of emotional input and further regulates cognitive or behavioral outcomes (Kalafatakis et al., 2018).

There are many other forms of ultradian rhythms in diverse physiological functions. Sleep, a representative circadian phenomenon, is also composed of an ultradian rhythm of rapid eye movement (REM)/non-REM (NREM) sleep cycles that have a period of about 90 min (Lamont and Amir, 2010). Ultradian sleep cycles are established during early childhood, and the duration of the ultradian REM/NREM cycle increases with age to a more adult-like pattern (Lopp et al., 2017). In a special case, oxytocin neurons in the hypothalamus display synchronized bursting electrical activity during lactation, every 3 to 15 min to induce milk ejection (Israel et al., 2003). As another example, ultradian locomotor rhythms in voles rely on the dopamine ultradian oscillator (DUO), which cycles in harmony with the circadian clock, and secreted dopamine levels fluctuate in synchrony with ultradian activity cycles (Blum et al., 2014; Bourguignon and Storch, 2017). Furthermore, calcium oscillations in cells can represent signals to transmit biological information that can be modulated by frequency or amplitude (Smedler and Uhlen, 2014). Synchronized calcium signals in hypothalamic slices of the subparaventricular zone and paraventricular nucleus (SPZ-PVN) showed an ultradian rhythm that was superimposed on the circadian rhythm in the SCN (Wu et al., 2018).

To elucidate the origin of the GnRH pulse generator, mathematical modeling and experimental confirmation were performed to modulate luteinizing hormone secretion by activating kisspeptin neurons (Voliotis et al., 2019). Thus, methods to measure ultradian rhythms exist, and research to find the mechanisms underlying ultradian rhythms is ongoing. Meanwhile, the hypothalamic ARN consists of a heterogeneous neural population including kisspeptin neurons. Kisspeptin neurons are critical regulators of mammalian reproduction. Mutations in the kisspeptin receptor caused hypogonadotropic hypogonadism in humans and mice, suggesting that its signaling is essential for puberty (Seminara et al., 2003). Consequential studies supported that kisspeptin neurons are upstream regulators of GnRH neurons and the hypothalamic-pituitary-gonadal axis that govern reproductive maturation (Pinilla et al., 2012). Kisspeptin neurons are immunohistochemically detected from embryonic day 16.5, and by that time they are synaptically connected to GnRH neurons (Kumar et al., 2015), unlike another kisspeptin population in the anteroventral periventricular nucleus (AVPV), which is detected at later developmental stages (Clarkson and Herbison, 2006). Previously, our group reported that intermittent administration of kisspeptin activated pulsatile GnRH transcription and secretion (Choe et al., 2013). GnRH secretion is observed soon after birth in the neonatal period (sometimes called the minipuberty of early infancy), then, after a quiescent juvenile period, reactivated from the peripubertal period (Fig. 1A), and the activity of GnRH is thought to involve the input from kisspeptin neurons (Herbison, 2016). During the follow-up study, we examined whether ARN kisspeptin neurons can generate pulsatility in the organotypic culture of neonatal mice. Kisspeptin neuron-specific calcium imaging from organotypic slice cultures revealed that ARN kisspeptin neurons generate self-sustained and synchronized calcium oscillations with an approximately 3-min period ex vivo (Kim et al., 2020).

In order to explore the factors that are involved in regulating these unique calcium oscillations, various possibilities were pharmacologically investigated (Fig. 1B). Voltage-gated sodium or potassium channel blockers disturbed calcium oscillations, indicating that action potentials are critical for generating oscillations. Chemogenetic inhibition using designer receptors exclusively activated by designer drug (DREADD) systems suggested that network interactions within ARN kisspeptin neurons are involved in mediating synchronized oscillations. Notably, signaling through neurokinin B (NKB) or dynorphin (Dyn), which is known to auto-regulate ARN kisspeptin neurons (also called KNDy neurons), was marginally effective at the neonatal stage. Furthermore, synchronized calcium oscillations were dependent on intracellular calcium regulation in the endoplasmic reticulum, but were not regulated by gap junctions composed of connexins 36, 50, or 43. Finally, N-methyl-D-aspartate (NMDA) receptor- and gamma-aminobutyric acid (GABA) type A receptor-mediated neurotransmission were involved in regulating ultradian calcium oscillations (Kim et al., 2020).

In organotypic slice cultures, kisspeptin neurons maintained their calcium rhythmicity depending on the network interaction within the ARN. Analysis of hypothalamic ARN and median eminence identified distinct clusters of neural cells (Campbell et al., 2017). The ARN is composed of heterogenous populations as shown in the clustering results of single cell RNA-seq (Fig. 2A). Among the populations, the satiety and hunger sensing neurons proopiomelanocortin (POMC) and agouti-related peptide (AgRP) which play opposite roles in energy homeostasis, are present that can be regulated by kisspeptin, suggesting a functional relationship between energy homeostasis and reproduction within the ARN (Fu and van den Pol, 2010). Ghrelin, a ‘hunger’ hormone released from the gastrointestinal tract, shows ultradian fluctuations (Yildiz et al., 2004) and increases the activity of AgRP/NPY while inhibiting that of POMC (Muller et al., 2015). Another population, dopaminergic neurons that inhibit prolactin release, are also regulated by kisspeptin (Ribeiro et al., 2015; Szawka et al., 2010). In particular, as mentioned in the previous section, DUO operates in the mammalian brain (Blum et al., 2014; Bourguignon and Storch, 2017). Collectively, these lines of evidence suggest that neuronal subtypes of ARN may coordinate kisspeptin neurons’ calcium oscillations through local neuronal interactions. We believe that future studies on these neuronal networks need to consider the heterogeneity of the ARN population.

ARN kisspeptin neurons can generate ultradian calcium oscillation early on, from the postnatal stage, which can possibly reshape with development. Fiber photometry of ARN kisspeptin neurons in adult mice exhibited calcium oscillations with a period of approximately 9 min (Clarkson et al., 2017), which is longer than the oscillation in neonates ex vivo (3 min). Between neonate and adult mice, numerous developmental changes take place (Fig. 2B). For example, the expression levels of the neuropeptides’ receptors implicated in the auto-regulation of KNDy neurons are low in neonates (Kim et al., 2020) and increase with developmental maturation (Gill et al., 2012; Navarro et al., 2012). The physiological significance of these short calcium oscillations remains to be elucidated.

Ultradian rhythms are also present in many phenomena during developmental processes. The basic helix-loop-helix transcription factors Achaete-scute homolog (Ascl1) and Hairy and enhancer of split-1 (Hes1), which regulate cell proliferation and differentiation, are expressed in an oscillatory manner to maintain proliferating neural progenitor cells, while sustained expression promotes neuronal fate determination (Imayoshi et al., 2013). Hes1 transcription is repressed by Hes6, a transcriptional regulator that shows rhythmic expression mediated by sumoylation (Lee et al., 2015). Additionally, Wnt and Notch signaling oscillations are functionally linked during segmentation of mouse embryonic mesoderm, and anti-phase oscillations impair physical segment formation (Sonnen et al., 2018).

In the developing embryo, the presomitic mesoderm is segmented into somites in a periodic manner by rhythmic gene expression along the antero-posterior axis (Kageyama et al., 2009). The periodic formation of somites is evolutionarily conserved across several vertebrate species (Wolpert et al., 2015): 90 min in chick, 120 min in mouse, 45 min in Xenopus, 30 min in zebrafish, and 6 h in humans. These oscillations in the ultradian time domain are driven by the segmentation clock, which is composed of periodic signaling activities and gene expression (Hubaud and Pourquie, 2014). Cycling of the Notch, Fgf, and Wnt signaling pathways as well as alternation between activation and repression of Hes and Hes-related factor expression underlies the somite clock, sharing the principle of the circadian rhythm. Somite formation is determined by the clock and wavefront model with signaling gradients. The presomitic mesoderm is committed to a specified pattern, as rotating a portion results in the formation of inverted somites. A surgical experiment manipulating the arrangement of somites demonstrated that the somite clock is functional within a given period of development, providing another clear example of the developmental control of ultradian rhythms (Wolpert et al., 2015).

Likewise, it is important that transcriptional factors involved in the developmental processes oscillate with a proper range of time. When the unstable Hes7 protein was stabilized to have a longer half-life, without changing the repressor activity, the mouse somite segmentation was disorganized (Hirata et al., 2004). Notably, the delay between expressions was caused by the introns of the Hes7 allele. In mice bearing the Hes7 locus with introns removed, somites were not properly segmented, and the segmentation defects in Hes7 null mice were rescued by introducing transgenes containing Hes7 introns (Takashima et al., 2011). Now, with the advance of techniques such as optogenetics, methods to modulate ultradian oscillations have become more precise. Therefore, interpreting the dynamics with spatiotemporal accuracy and examining the functional role of ultradian oscillations will expand our understanding of ultradian rhythms in developmental processes (Isomura and Kageyama, 2014).

We reviewed the characteristics of ultradian rhythms with special reference to the calcium oscillations in hypothalamic ARN kisspeptin neurons and developmental processes. Ultradian rhythms are prevalent in living organisms and can affect many physiological systems, including reproductive and developmental processes. As was the case for the circadian rhythm, revealing the mechanisms of the ultradian rhythm will fundamentally facilitate the understanding of ultradian biological processes, as well as strategies to modulate them. One of the difficulties in studying the ultradian rhythms is that they can be driven by either genomic or non-genomic mechanisms, or both. In the somite clock, a negative feedback loop incorporating transcription and translation, in combination with the cyclic activation of signaling pathways, plays a key role in its oscillation, supporting the idea of a genomic drive. Gene expression is also crucial in mediating the GnRH pulse generator and its pulsatile responses to intermittent kisspeptin treatment (Choe et al., 2013). In contrast, the calcium oscillation of ARN kisspeptin neurons is strongly regulated by ion channels and inputs via NMDA receptors. Ultradian calcium oscillation found in the SPZ-PVN region is well controlled by glutamatergic transmission. Thus, both genomic and non-genomic mechanisms are involved in regulating ultradian rhythms.

Another difficulty is that the ultradian rhythm is often overruled by circadian oscillation, complicating the analysis of time-series biological data. We speculate that technical advances in bioinformatics and genome-wide approaches will accelerate the understanding of ultradian rhythms manifested in complex biological systems. Indeed, in silico analysis of the hepatic transcriptome identified the expression of hundreds of ultradian genes, and key DNA motifs in the promoter regions of these genes (van der Veen and Gerkema, 2017). Thus, further studies should focus on both genomic and non-genomic mechanisms that drive ultradian oscillations to understand ultradian rhythms, the fundamental form of biological oscillation that regulates developmental and physiological processes.

This work was supported by the Ministry of Science, ICT and Future Planning through the National Research Foundation of Korea (NRF-2017R1A2A1A05001351) and DGIST Start-up Fund Program (2020010055).

D.K. conceived and wrote the manuscript. H.K.C. consulted the design and provided feedback. K.K. supervised, provided expertise, and secured funding.

The authors have no potential conflicts of interest to disclose.