Limitations and perspectives

Coupling in various model systems

U-2 OS cells constitute one of the most commonly used in vitro models in chronobiological research due to their (i) human origin, (ii) extensive characterization, (iii) easy handling, (iv) stable oscillations, and (v) susceptibility to genetic manipulations. Nevertheless, it has to be kept in mind that this cell line is derived from cancerous tissue with aberrant genetic material and behavior. Especially TGF-b

signaling is often deregulated in cancer, leading to altered signaling activity or responsiveness of tumor cells [443], [444].

Our findings suggest that that paracrine communication mechanisms are conserved across human and murine species, as well as across a number of peripheral tissues.

Additionally, published studies in non-transformed human, murine, and even zebrafish model systems have yielded evidence of interactions between the molecular circadian clock machinery and TGF-b signaling pathway [335], [336], [339], [340]. Nevertheless, identification and functional role of TGF-b as paracrine signaling molecule and potential peripheral coupling factor based on the U-2 OS cell model should interpreted carefully until further validation. Nevertheless, despite drawbacks of in vitro culture systems, coupling constitutes an integral feature of cellular oscillators. Perturbation of coupling on the cellular level can affect the behavior of entire oscillator networks, e.g.

its rhythmic biological functions [445], entrainment [57], or response to non-rhythmic signals [446]. Thus, even though cell-based systems constitute simplified and isolated models of complex in vivo phenomena, in vitro studies mark an important starting point for understanding molecular mechanisms of peripheral coupling.

As discussed, ex vivo 3-dimensional culture models seem to reflect in vivo tissue configurations and may provide new insights into coupling among peripheral circadian oscillators. Therefore, organoids derived from mammalian stem cells may be a good model for studying the role of TGF-b in peripheral coupling independently of its role in cancerous processes. In vivo, peripheral clocks have been shown to oscillate independently of the SCN, behavioral rhythms, external light-dark and feeding-fasting cycles [36]. However, whether or not tissues rhythms are maintained by intercellular coupling or by interaction with other peripheral tissue clocks needs to be elucidated.

Neither in vitro nor ex vivo models can provide information about such complex processes. However, the role of TGF-b signaling for intercellular coupling within peripheral tissues, as well as for rhythmic organ functions can be studied in vivo. Newly developed mouse models [244], [245] and imaging techniques [36] allow for real-time recording of bioluminescence rhythms of isolated peripheral tissues, i.e. in otherwise clock-less animals. Thus, amplitude and damping parameters of free-running tissue rhythms, as well as response to Zeitgeber/entrainment signals with or without functional TGF-b signaling can be investigated in living animals. Moreover, functional consequences of disturbed TGF-b signaling in isolated peripheral clocks, as well as in

clocks receiving systemic and external Zeitgeber signals may be studied to elucidate the role of peripheral coupling for the temporal coordination of circadian tissue physiology.

Population versus single cell imaging

Population imaging, i.e. quantification of the average rhythm of cellular ensembles, for studying intercellular coupling constitutes one of the major limitations of this project.

Circadian networks are composed of cell-autonomous single cell oscillators, which all cycle with their individual circadian parameters (period, phase, amplitude, damping).

Thus, generally, intercellular coupling strength, which itself is difficult to quantify, is approximated by the distribution of circadian parameters of single cell oscillators. The widths of these distributions reflects the degree of synchronization within the network [172]. However, changes in amplitude, phase, and period distributions may also result from changes of cell-intrinsic oscillations independently of coupling. Therefore, population averages, as presented here, cannot clearly distinguish between changes of individual oscillators, changes of intercellular coupling between oscillators or a combination of both. Nevertheless, damping is commonly accepted to reflect desynchronization rather than damping of individual oscillators [48], [62], [447], [448].

Additionally, bidirectional phase-/period-pulling effects, as well as amplitude expansion upon physical separation of co-cultures are unlikely to arise simply from changes of single cell oscillators alone. Thus, even without single cell imaging, our data as whole strongly suggests that observed changes in circadian parameters depend on intercellular coupling or a combination of single cell changes and coupling.

In the context of cell division, precise separation between single cell and population-based effects may be of importance when studying intercellular coupling. O’Neill and Hastings (2008) suggest that desynchronization of fibroblasts is decreasing as cells mature because “phase-noise”, introduced by cell division, decays [355]. Other studies have demonstrated unidirectional influence of cell cycle on circadian cycle resulting in period changes of circadian oscillations [47], [449], [450]. Considering that intercellular coupling is achieved by and frequency-locking between oscillators, phase-/period-fluctuations introduced by cell division may lead to undetectable artifacts when studying coupling on the population level. However, as many in vitro models, U-2 OS cells exhibit contact inhibition under dense culture conditions. Thus, for most results

presented here, cell cycle effects can be neglected. Moreover, bioluminescence imaging was conducted using serum-free reporter medium in order to arrest also sparse cultures in G0/G1 phase and mimic in vivo situations of non-dividing resident tissue cells. TGF-b is known as cell cycle regulator, inducing G1 arrest in most cell types, e.g. epithelial, hematopoietic and endothelial cells, but promoting growth of certain mesenchymal cells such as skin fibroblasts [336]. We have to admit that by population imaging we cannot exclude that TGF-b dependent cell cycle regulation impacts intercellular coupling. Especially pharmacological perturbation of TGF-b signaling may introduce growth effects that cannot be separated from coupling. Thus, even though we suggest that intact TGF-b signaling is promoting synchrony by peripheral oscillator coupling, we are aware that additional single cell recordings are necessary to strengthen our findings. For example, single cell imaging with dual reporter systems may enable to study phase dispersion of peripheral oscillators (as measure for coupling strength), as well as cell cycle progression upon perturbation of TGF-b signaling simultaneously. Alternatively, without a cell cycle reporter system, growth effects of TGF-b receptor inhibition could be controlled by quantifying cell number and/or cell covered surface area of times series data.

Technical and experimental limitations

RNA interference (RNAi) screens constitute an error prone technique. As for every high-throughput RNAi screen, short-hairpin RNA (shRNA) knock-down efficiencies could not be controlled by quantification of transcript and/or protein levels. Comparable lentivirus delivery-based gene silencing techniques often result in decreases of mRNA levels to < 25%. However, during a screen, it cannot be excluded that shRNA mediated silencing fails completely, that knock-down efficiencies are too low to disturb functionality of the targeted gene, or that off-target effects are introduced. Thus, as mentioned, results from RNAi based screens should be interpreted with care and the role of identified target genes should further be validated by more reliable genome editing approaches, e.g. CRISPR/Cas9 gene editing or generation of transgenic animal. So far generation of mutated/knock-out cells, such as SMAD4 or TGF-b receptor depleted U-2 OS cells, has not been done but should be attempted in the future. This would allow to study intercellular coupling without being dependent on pharmacological inhibitors, which are often accompanied by unspecific effects.

Differential gene expression (DGE) analysis and mass spectrometry are usually accompanied by technical constraints with regard to bioinformatical analysis of resulting datasets. While mass spectrometry may fail to detect very low abundant peptides in a complex mixture of proteins or to separate isoforms, DGE analysis may overestimate statistical relevance of expression changes of lowly expressed genes.

Moreover, for both experimental approaches, transcripts and peptides have to present in available databases to be identified. We did the best to minimize these error sources by enriching active conditioned medium factors by chromatography prior to mass spectrometry, as well as by filtering and normalization of RNA sequencing data prior to DGE analysis (according to [350], [351], [451]). It appears unlikely, that for the human genome functional transcripts or peptides are not annotated.

The lack of evidence for a direct connection between TGF-b dependent CRE activation and modulation of the molecular clock machinery, as well as the observed phase responses constitutes an additional limitation of this project. Kon et al. (2008) have shown that TGF-b stimulation alters clock gene expression levels and elicits time-of-day dependent phase responses [339]. Moreover, it has been shown that rhythmic transcription factor binding to CRE sites in the Period promoter is important for oscillations of this gene and circadian rhythmicity in peripheral oscillator models [395].

Our results suggest that TGF-b signaling promotes synchronized and robust network oscillations. However, whether SBE/CRE-driven PER2/Per2 induction constitutes the input pathway of TGF-b signals to the circadian clocks remains to be studied in detail.

We suggest that coupling experiments in cells/tissues with mutated or depleted clock-relevant CRE sites in the endogenous Period gene promoter should be conducted. If TGF-b dependent CRE induction of PER2/Per2 is required for intercellular coupling and/or phase responses these effects should be lost upon genetic manipulation.

Additionally, whether TGF-b signaling induces CRE-driven Period transcription could be measured by ChIP sequencing using aSMAD/aCREB antibodies or proximity labeling of transcription factors present at Per2 CRE site following TGF-b stimulation.