The molecular mechanism underlying the periodic induction of lateral roots, a paradigmatic example of clock-driven organ formation in plant development, is a matter of ongoing, controversial debate. Here, we provide experimental evidence that this clock is frequency modulated by light and that auxin serves as a mediator for translating continuous light signals into discontinuous gene activation signals preceding the initiation of lateral roots in Arabidopsis seedlings. Based on this evidence, we propose a molecular model of an ultradian biological clock involving auxin-dependent degradation of an AUX/IAA-type transcription repressor as a flexible, frequency-controlling delay element. This model widens the bandwidth of biological clocks by adding a new type that allows the pace of organ formation to adapt to the changing environmental demands of the growing plant.
In contrast to animals, the development of plants is characterized by a life-long ability to produce new organs, such as stem internodes, branches or leaves in a repetitive pattern. The temporal spacing of these events can vary from a few hours to many days, and depends on environmental conditions, such as light or nutrient supply, in addition to intrinsic cues (Malamy, 2005; Tian et al., 2014). In this respect, recurrent organ formation differs fundamentally from phenomena controlled by the stabilized circadian rhythmicity governed by endogenously regulated clocks. In contrast to circadian clocks, our understanding of the rhythmic processes with variable, regulated periodicity is currently limited. However, periodic de novo formation of lateral roots (LR) along the primary root of Arabidopsis thaliana seedlings has recently emerged as a promising experimental system for tackling this problem (Péret et al., 2009; Van Norman et al., 2013; Du and Scheres, 2018).
The root tip region where the initiation of LR primordia (LRP) in the model plant Arabidopsis normally occurs (designated here as the primordium formation zone) is partitioned into functionally specialized regions: the root apical meristem (0.3-0.4 mm in length), basipetally (shootward) followed without a sharp boundary by the elongation zone (∼0.7 mm), followed by the maturation zone up to the earliest detectable LRP (∼5 mm). Under homeostatic growth conditions, the root apical meristem produces new cells with a constant rate and delivers them to the elongation zone, where they elongate up to 4-fold within 2 h before being added to the maturation zone (Verbelen et al., 2006). Tied up with this continuous flow of cells along a polarized ‘assembly line’ is the discontinuous generation of lateral meristems from certain pericycle cells giving rise to LRP in the early maturation zone in a regular periodic fashion. The initiation of formative cell divisions in LR founder cells is preceded by the generation of a prepattern of gene activation that can be visualized by the expression of auxin-responsive pDR5-directed reporter genes (De Smet et al., 2007; Moreno-Risueno et al., 2010). Originally, the first sign of prepattern formation was detected as periodic expression of the auxin-responsive pDR5:GLUCURONIDASE (GUS) reporter gene in opposite protoxylem cell strands, followed by neighboring pericycle cells in the basal root apical meristem, designated the priming reaction (De Smet et al., 2007). Related observations in vivo with the pDR5:LUCIFERASE (LUC) reporter subsequently showed (Moreno-Risueno et al., 2010) that the priming reaction is accompanied, or followed, by a pulse of massive gene activation in the ‘oscillation zone’, with its center at the shootward end of the elongation zone. Prepattern formation is completed by the emergence of distinct, stable spots of pDR5:LUC activation, designated as prebranch sites (PBS), which mark the position of future LRP (Moreno-Risueno et al., 2010). In this context, these so-called ‘oscillations’ are in fact apparent, non-cell-autonomous oscillations (Laskowski and Ten Tusscher, 2017). The oscillation zone represents a continuously changing population of cells moving through this zone, where they can be imprinted with new gene expression identities. Under normal conditions of growth, at an elongation rate of ∼0.4 mm h−1 and intervals of a few hours, these periodic gene activation pulses hit particular cells only once on their transit through the oscillation zone. Thus, at the cellular level, such an ‘oscillation’ is a singular event, representing the singular output signal of an oscillator mechanism acting upstream of this event. In the kymographic presentation of PBS data (see below), the development of single, spatially separated pDR5:LUC pulses followed in time by continuous LUC expression tracks is clearly seen.
Current efforts aimed at elucidating the developmental program underlying the periodic prepattern formation in the primordium formation zone are characterized by a controversial debate over the role of the plant hormone auxin in the time-keeping mechanism regulating this process. Based on the periodic expression of the pDR5:GUS reporter in the basal root apical meristem, it has been hypothesized (De Smet et al., 2007) that periodic fluctuations in auxin distribution or responsiveness determine the longitudinal spacing of LR initiations. Inhibition of pDR5:GUS expression by PEO-IAA, an auxinole-type antiauxin (Hayashi et al., 2008; De Rybel et al., 2010), revealed that priming is mediated by the TIR1/AFB signaling pathway generally involved in auxin-regulated developmental processes in plants. However, because of the failure of applied natural auxin (indole-3-acetic acid, IAA) to affect the frequency of pDR5:LUC pulses in the oscillation zone and to produce expected expression patterns of some pAUX/IAA:LUC reporters, it was concluded (Moreno-Risueno et al., 2010; Van Norman et al., 2013) that the pace of PBS production is set by an endogenously regulated clock with a stable, auxin-independent periodicity. On first sight, this concept is difficult to reconcile with the familiar usage of auxin as a tool for promoting LR formation, and casts doubts upon the reputation of pDR5-directed reporter genes as reliable indicators of responsiveness to auxin. Analyzing the transformation of pDR5:LUC pulses into PBS in mutants impaired in the conversion of indole-3-butyric acid (IBA) into IAA led to the proposal (Xuan et al., 2015) that auxin comes into play by modulating the amplitude of clock-based oscillations, thereby determining the fraction of successful priming events. According to a more recent proposal (Xuan et al., 2016), periodic bursts of programmed cell death in the lateral root cap release pulses of auxin to underlying root tissues, thereby establishing the frequency of oscillations. This and other contrasting hypotheses for the generation of oscillations in auxin concentration or responsiveness have been discussed in detail elsewhere (Laskowski and Ten Tusscher, 2017). In summary, it appears that the presently available knowledge does not provide a clear picture that allows the role of auxin, if any, to be pinpointed in the mechanism regulating the periodicity of LR formation in the primordium formation zone. Thus, three questions appear to be of utmost concern and are addressed here: (1) is LR formation controlled by a clock with a stable, or a variable periodicity open for regulatory inputs by environmental factors such as light?; (2) is auxin part of the time-keeping mechanism dictating the stroke of the clock by instructive signals, or merely a necessary, permissive component of the signal output chain (Del Bianco and Kempinski, 2011; Van Norman et al., 2013)?; and (3) if the first applies, how can continuous light or auxin levels be converted into discontinuous signals, the frequency of which depends on light intensity or auxin concentration? The aim of the present contribution is to provide answers to these basic questions.
Although the ability of exogenous IAA to stimulate synchronous LR formation along the mature region of the primary root of light-grown Arabidopsis seedlings is well documented (Laskowski et al., 1995), corresponding experiments specifically addressing periodic LR formation in the primordium formation zone have rarely been reported. In fact, we are aware of only one paper (Ivanchenko et al., 2010) describing an increase in LR, correlated with an increase in primary root elongation, by IAA in the range of 1-5 nM in the tip region of slowly growing roots. Under the growth conditions used in the current study, 10-270 nM IAA eliminated LR formation in the primordium formation zone, correlated with severe inhibition of root elongation (Fig. 1A,C). However, replacing IAA with the synthetic auxin naphthalene-1-acetic acid (NAA) increased the frequency of LR formation in the primordium formation zone without affecting root elongation (Fig. 1A,C). A similar effect was produced by feeding plants with tryptophan (0.01-0.3 mM), a precursor of IAA biosynthesis (Zhao, 2012) (Fig. 2A,B).
In the absence of light, young seedlings kept on sucrose-containing medium are capable of producing short roots with 0-2 LR close to the root base (Bhalerao et al., 2002). Similar to light-grown seedlings, the roots of dark-grown seedlings were induced to exhibit periodic LR formation by NAA, but not by IAA (Fig. 1B). Tryptophan did not promote LR formation in dark-grown seedlings, suggesting that its conversion into IAA is a light-requiring process (Fig. 2A).
As depicted in Fig. 1, light has a dramatic effect on LR formation. Dose-response curves (Fig. 3A) obtained by transferring seedlings raised in standard light (130 µmol m−2 s−1) to a range of fluence rates of 0-250 µmol m−2 s−1 showed that increasing the light flux gradually increased the frequency of LR+LRP formed (from 1.2±0.1 day−1 to 6.4±0.6 day−1), as well as the fraction of emerging LR, accompanied by a 1.4-fold promotion of root elongation (Fig. 3B). Although there is no evidence that root growth changes in this range affect the frequency of LR formation (Moreno-Risueno et al., 2010), it cannot be excluded that root elongation has some influence. Important in the present context, these data do not take into account the occurrence of non-persistent PBS that cease development before producing detectable LRP (Moreno-Risueno et al., 2010). Counting the total number of nascent PBS mirroring the time pattern of oscillations revealed a dose-dependent effect of light on PBS frequency, exceeding the frequency of LR+LRP under standard light by 2.4-fold (Fig. 3C). An integral view of the temporal and spatial pattern of PBS formation can be obtained by transforming time-lapse imaging data (Fig. 3D,E; Movie 1) into kymographs (Xuan et al., 2015) (Fig. 3F). This kind of data presentation revealed the appearance of (sometimes weakly expressed) PBS that suspended pDR5:LUC activation after ≤24 h and that did not continue development even after several days in the light (Fig. 3D-F). The number and distribution of LR+LRP ultimately produced (6.2±0.1 day−1) corresponded to the number of persistent PBS, demonstrating continued pDR5:LUC activation (Fig. 3C).
The data presented so far provide evidence that: (1) the oscillator(s) responsible for adjusting the pace of PBS and LR formation can be frequency modulated by light; and (2) a variable proportion of the PBS can be selected for further development in a light-dependent manner while the rest becomes arrested. To further pinpoint the role of auxin during the early steps of LR formation, we examined the effect of NAA application on PBS formation in dark- and light-grown seedlings. In the absence of light, the roots exhibited irregular patches of non-persistent PBS, with occasionally one persistent PBS close to the root base (Fig. 4A,C). Low levels of NAA (30-100 nM) induced an increase in the frequency of total PBS and an increase in the fraction of persistent PBS that matched the increase in LR+LRP produced later (Fig. 4A,D). In the light, the relative effect of NAA on total PBS was smaller, presumably because the response approached saturation with respect to endogenous auxin (Fig. 4B). Again, NAA increased the production of persistent PBS and LR+LRP formed later to the same extent (Fig. 4B,E,F).
It was previously shown (Bhalerao et al., 2002) that light induces auxin accumulation in the root tip of Arabidopsis seedlings. To examine in more depth the dependence of LR formation on endogenous auxin production and perception in light-grown seedlings, we took advantage of a pharmacological approach that, in contrast to functionally related mutants, enabled us to modify auxin synthesis and perception in a graded manner in a defined developmental stage of plants with undisturbed genetic background. We used kynurenine (KYN), a competitive inhibitor of TAA1/TAR tryptophan aminotransferases involved in IAA biosynthesis from tryptophan (He et al., 2011), and auxinole (AXO), a competitive inhibitor of auxin perception at the TIR1/AFB receptors (Hayashi et al., 2012). Both KYN and AXO attenuated the frequency of LR+LRP production in a concentration-dependent manner, with little effect on root elongation (Fig. 5A-D). Oscillatory priming and PBS generation was inhibited by >90% with 10 µM KYN or AXO (Fig. 5E-H). The NAA-reversible (Fig. 5A,C, insets) inhibition by KYN indicates that the IAA controlling LR initiation results from the tryptophan pathway. Thus, although applied IBA promoted LR formation as previously reported (Xuan et al., 2015) and this effect was even enhanced by KYN (Fig. S1), a significant involvement of endogenous IBA-derived auxin (Xuan et al., 2015, 2016) could not be confirmed in our experiments. The inhibitory effect of AXO was also shown in dark-grown seedlings (Fig. 5B,D, insets). In agreement with its antiauxin function, the concentration of AXO needed to counteract NAA varied with the concentration of NAA in an interactive manner (Fig. S2). LR+LRP formation was inhibited by KYN or AXO applied to the growing root tip (Fig. S3), demonstrating that auxin synthesis and perception involved in LR initiation occur in the primordium formation zone.
Taken together, our data provide evidence that: (1) the timing mechanism of LR formation is controlled by light (Figs 1-4); (2) the oscillator(s) responsible for adjusting the pace of PBS and LR formation can be frequency modulated by up- and downshifts in global auxin concentration (Figs 4 and 5A,E); and (3) auxin is necessary and responsible for mediating the effect of light on the timing of PBS generation, as well as for recruiting PBS for further development into LRP (Figs 4 and 5). This suggests that periodic generation of PBS and their periodic selection for continued development are distinct clock-controlled processes. However, this point requires further experimental clarification. Incidentally, our results invalidate possible doubts concerning the qualification of pDR5-directed reporters as reliable indicators of auxin activity.
In agreement with a previous report (Moreno-Risueno et al., 2010), Fig. 1 shows that exogenous IAA (in contrast to exogenous NAA) appears unsuitable for experimentally demonstrating the LR-initiating function of auxin in the primordium formation zone of the root. However, other than exogenous IAA, applied IAA precursor tryptophan promoted LR formation in an AXO-inhibitable manner (Fig. 2A), providing supportive evidence that IAA can mediate LR formation in the primordium formation zone if produced by an endogenous pathway. In accordance with these results, it was previously shown (Xuan et al., 2015) that the application of the IAA precursor IBA increases the frequency of the oscillations as well as PBS and LR formation. These apparently contradictory findings are reminiscent of the failure of exogenous IAA in rescuing auxin deficiency in yuc IAA biosynthesis mutants, whereas the introduction of a bacterial biosynthesis gene under the control of a YUC promoter restores the phenotype (Cheng et al., 2006). Taken together, these data show that auxin has an important role in periodic LR formation in the primordium formation zone.
Transient DR5:LUC pulses vanishing before primordia are formed have been demonstrated previously (Moreno-Risueno et al., 2010; Xuan et al., 2015). Detailed quantitative analysis of mutants with diminished auxin perception or IBA-to-IAA conversion revealed that the number of PBS and LRP formed was lower than expected from the number of oscillations and was correlated with a decrease in oscillation signal intensity rather than in oscillation number (Xuan et al., 2015). This finding led to the conclusion that auxin modulates the amplitude of the priming reaction, thereby determining whether a persistent PBS is created (Xuan et al., 2015). Here, we showed that a similar relationship can be demonstrated in the case of non-persistent and persistent PBS. As a general result from these observations, the frequency of priming pulses or incipient PBS and the frequency of the subsequently resulting LRP can differ significantly because of the light-dependent arrest of a variable proportion of PBS at an early stage of development. According to a previous proposal (Xuan et al., 2015), the arrested stages could represent a stock of silent precursors for later usage, such as after damage to the primordium formation zone. Moreover, they could be involved in the synchronous induction of LR (Laskowski et al., 1995) by high auxin concentrations in the maturation zone (Van Norman et al., 2013). Further work is needed to test these hypotheses.
Our findings emphasize that the signaling events preceding LR initiation are highly flexible and can be modulated by the active auxin level in darkness and light. This raises questions as to the molecular mechanism of the oscillator(s) involved and the role of auxin in permitting their flexible performance. The auxin-signaling pathway of developmental responses, such as LR formation, involves the perception of the hormone by the TIR1/AFB receptors, initiating the ubiquitin-mediated degradation of canonical members of the family of AUX/IAA repressor proteins, which otherwise prevent certain ARF transcription factors from activating auxin-inducible target genes (Salehin et al., 2015). Given that these genes include those encoding AUX/IAAs, the degradation of AUX/IAAs also promotes their own synthesis, which, upon return to an inhibitory level, shuts down their transcription by means of a negative feedback loop (Hagen and Guilfoyle, 2002). It appears to be predominantly assumed that LR formation is triggered by the periodic establishment of spots of auxin accumulation (‘auxin maxima’) at the sites of future LRP initiation (De Smet et al., 2007; Dubrovsky et al., 2008; Péret et al., 2009; De Rybel et al., 2010; Del Bianco and Kepinski, 2011,; Laskowski, 2013; Xuan et al., 2016; Taylor-Teeples et al., 2016; Laskowski and Ten Tusscher, 2017). This view has mainly been based on the local expression of pDR5-directed reporter genes serving as a proxy for ARF-inducible genes and, thus, as an indirect and, therefore, ambiguous proxy for auxin accumulation (Del Bianco and Kepinski, 2011). However, how these postulated auxin maxima could be generated under conditions of continuous auxin synthesis or uptake remains a matter of conjecture (Laskowski and Ten Tusscher, 2017). Moreover, this idea presupposes the existence, rather than reveals the mechanism, of the underlying periodicity-generating process. In the case of the membrane-permeable auxin NAA, carrier-mediated influx can be excluded as a means to modify intracellular auxin levels (Delbarre et al., 1996).
Consistent with results obtained with exogenous IBA (Xuan et al., 2015), our data (Figs 3-5) showed that LR formation is preceded by a periodic production of PBS with time intervals that can be modified by increasing or decreasing the stationary level of, or the receptor accessibility for, auxin in the responsive tissue. Thus, any model explaining these findings must rely on an oscillatory mechanism capable of converting, at a post-perception stage, time-invariant auxin signaling into a temporal sequence of distinct all-or-nothing responses, the frequency of which depends on auxin concentration. To develop a model that satisfies this crucial requirement, we chose a perspective differing from the prevailing view by focusing on the dynamics of an auxin-sensitive AUX/IAA-type ARF repressor instead of on local auxin distribution. Experimental evidence indicates that AUX/IAA proteins can be degraded in vivo in a first-order reaction, with half-lives on the order of minutes or hours, depending on the ambient concentration of auxin (Kepinski and Leyser, 2002; Dreher et al., 2006). Recapitulation of the Aux/IAA-TIR1/AFB-ARF signaling pathway in yeast revealed that the degradation rate of Aux/IAA proteins determines the transcriptional activity of related ARF transcription factors (Pierre-Jerome et al., 2014). In plants expressing artificially generated degradation variants of IAA14, an AUX/IAA protein involved in LRP initiation (Fukaki et al., 2002), LR formation in the growing root was correlated with the rate of auxin-dependent IAA14 degradation (Guseman et al., 2015). It was concluded that Aux/IAA proteins can function as auxin-tuned timers for synchronizing developmental transitions (Guseman et al., 2015). Based on these results, we replace the ‘auxin-maximum concept’ with an ‘AUX/IAA-minimum concept’, stating that pulses of low, inactive levels of a special AUX/IAA protein (designated as <AUX/IAA>), interrupted by periods of higher, inhibitory <AUX/IAA> levels, represent the trigger events releasing a suitable partner ARF (<ARF>), and therewith LR-related target genes, including DR5:LUC, from repression. In line with this idea, a mechanistic model of an oscillator regulating the frequency of LR initiation without involving fluctuating auxin maxima can be derived as follows. A minimum model of a biological oscillator system comprises a promoter element and a repressor element arranged in a cycle involving a negative feedback loop (Novák and Tyson, 2008). If the reactions connecting these elements are sufficiently separated by a time delay to prevent them from settling at a steady-state, sustained oscillations can occur (Novák and Tyson, 2008). Provided that the delay depends at least in part on the half-life of the repressor, the oscillation period can be modulated by the rate of repressor sequestration (Novák and Tyson, 2008). As pointed out previously (Pierre-Jerome et al., 2013), a paradigmatic molecular example of this kind of oscillator mechanism is the NF-κB signaling system of mammalian cells, which generates ultradian oscillating changes in the activity state of an transcription factor in response to certain stress stimuli (Sung et al., 2009). Evidently, the AUX/IAA-TIR1/AFB-ARF signaling pathway potentially contains all the necessary components of a minimum model of an oscillator system: an activating promoter (<ARF>); a degradable repressor (<AUX/IAA>), the production of which depends on promoter activity; and a negative feedback loop connecting the repressor and promoter. Thus, it suggests that this pathway is a closed cycle with the ability to exhibit the oscillatory behavior of a biological clock (Middleton et al., 2010) (Fig. 6). Auxin can feed information into this circuit by binding to TIR1/AFB, thereby promoting the depletion of <AUX/IAA> and, thus, activating the transcription of auxin-inducible target genes, including <AUX/IAA>. Importantly, by including auxin-dependent <AUX/IAA> degradation, this cycle contains a variable delay element that enables the modulation of cycling time and, hence, signal-output frequency, as a function of auxin concentration. Mathematical modeling (Middleton et al., 2010) has shown that this kind of circuit can either operate in a steady-state mode or accomplish cell-autonomous oscillations under a constant auxin supply, depending on choosing the appropriate parameter regimes. In fact, it can intuitively be predicted that the oscillation comes to a halt if, at subcritical auxin concentrations, the <AUX/IAA> pool can no longer be reduced to a level allowing <ARF> activation. By contrast, if, at high auxin concentrations, the degradation rate of <AUX/IAA> exceeds the capacity of its synthesis, the oscillation becomes arrested in the active <ARF> phase. In agreement with these predictions, the entry of pericycle cells into the maturation zone is accompanied by the stopping of further LR initiation in these cells. However, excessive synchronous LR production can subsequently be induced in the maturation zone by application of high concentrations of auxin (Laskowski et al., 1995). In this case, the clock mutates into an OFF/ON switch set in motion by abrupt changes in stationary auxin signal strength. This mode of action could also apply to other cases where auxin mediates non-rhythmic responses, such as by maintaining high gene expression levels in PBS and later LRP stages. By tuning the degradation rate of <AUX/IAA> between these extremes (at suitable <AUX/IAA> synthesis), auxin can take over the function of a timing-signal provider (Guseman et al., 2015) for setting the stroke of a dynamic clock that adapts the pace of organ development to the actual demands of the growing plant exposed to variable, non-predictable environmental conditions. This decisive feature discriminates the LR induction clock from endogenous clocks with fixed periodicity (Moreno-Risueno and Benfey, 2011) used for executing fixed segmentation programs in animal embryo development or synchronizing behavior with the phasing of the external world.
So far, the model depicted in Fig. 6 is supported in qualitative terms by the data presented above. It now needs to be experimentally validated in detail by identifying the particular AUX/IAA and ARF species acting as oscillator core proteins, and by testing whether the reaction constants and time delays quantitatively satisfy the kinetic requirements dictated by the output response.
MATERIALS AND METHODS
All experiments were performed with pDR5:LUC-expressing transgenic A. thaliana seedlings (Col-0 ecotype) obtained from the Benfey lab (Moreno-Risueno et al., 2010). Seedlings were grown on sterile vertical agar plates without or with 10 g l−1 sucrose (as indicated) in continuous white fluorescent light (standard fluence rate 130 µmol m−2 s−1) or darkness at 24.0±0.5°C as previously described (Kircher and Schopfer, 2016). Green safelight was used for manipulating dark-grown seedlings, the germination of which was induced by 10 min light. Root regions of interest were marked for microscopic evaluation on the backside of plates. To minimize the photodegradation (Stasinopoulos and Hangarter, 1990) of light-sensitive chemicals (IAA, AXO and KYN), seedlings were shielded from the light below the hypocotyl in the relevant experiments by encasing the lower part of the plates in light-tight envelopes. In control experiments, root elongation and LRP formation were not significantly affected by shielding the roots from the light (Kircher and Schopfer, 2016). AXO was synthesized according to a published protocol (Hayashi et al., 2012). Fluence rate was attenuated by covering the (exclusively light-exposed) upper surface of vertically placed plates with 1-4 layers of fine metal gauze. To visualize LRP, roots were stained with acetocarmine (Kircher and Schopfer, 2016). PBS expressing LUC activity were visualized by time-lapse imaging of seedlings sprayed with 2 mM luciferin by using a Vers Array XP camera system (Roper Scientific) operated at exposure times of 5 min (binning 2) interrupted by 15 min darkness or 14 min LED light (standard fluence rate 130 µmol m−2 s−1; roots shielded from light) and 1 min darkness (Kircher and Schopfer, 2016). Image sequences were processed using FIJI/ImageJ software (Schindelin et al., 2012) and converted into kymographs by carefully selecting the final course of root growth using the ‘Segmented Line’ tool and applying the ‘Reslice’ function. Counting of PBS was performed visually by identifying clearly distinguishable spots with higher luminescence intensities compared with the local background. Adjustments of brightness and contrast (intensity scaling) were performed with equal values to allow comparability, except where noted.
Data points represent means±their estimated standard deviations (s.e.m.) of 4-6 independent experiments comprising 4-10 seedlings; see figure legends for details. Qualitative imaging data represent typical examples from at least four independent experiments.
We thank Dr P. Benfey and his group (Duke University, Durham, NC, USA) for providing seed materials and Dr W. Seiche (Chemistry Department, University of Freiburg) for synthesizing auxinole.
Conceptualization: S.K., P.S.; Methodology: S.K., P.S.; Validation: S.K., P.S.; Formal analysis: S.K., P.S.; Investigation: S.K., P.S.; Writing - original draft: P.S.; Writing - review & editing: S.K.; Visualization: S.K., P.S.; Project administration: S.K., P.S.; Funding acquisition: S.K., P.S.
This work was supported by the Deutsche Forschungsgemeinschaft (SCHO 154/8-1 to P.S. and KI 1077/2-1 to S.K.).
The authors declare no competing or financial interests.