|Home | About | Journals | Submit | Contact Us | Français|
Users may view, print, copy, download and text and data- mine the content in such documents, for the purposes of academic research, subject always to the full Conditions of use: http://www.nature.com/authors/editorial_policies/license.html#terms
Many protein kinases are key nodal signaling molecules that regulate a wide range of cellular functions. These functions may require complex spatiotemporal regulation of kinase activities. Here, we show that Protein Kinase A (PKA), Ca2+ and cAMP oscillate in sync in insulin-secreting MIN6 β cells, forming a highly integrated oscillatory circuit. We found that PKA activity was essential for this oscillatory circuit, and was capable of not only initiating the signaling oscillations but also modulating their frequency, thereby diversifying the spatiotemporal control of downstream signaling. Our findings suggest that exquisite temporal control of kinase activity, mediated via signaling circuits resulting from cross-regulation of signaling pathways, can encode diverse inputs into temporal parameters such as oscillation frequency, which in turn contributes to proper regulation of complex cellular functions in a context-dependent manner.
Protein kinases serve as major information transducers in cells and are responsible for regulating diverse cellular functions1,2. However, the molecular mechanisms underlying this signaling diversity are not well understood, particularly with regard to the role of the spatiotemporal regulation of kinases. Although important progress has been made in increasing our understanding of the spatial compartmentation of kinases3,4, much less is known about how temporal regulation of kinase action can be exploited to encode diverse signals and control different functional outcomes5.
PKA, a prototypical protein kinase6, plays a host of important roles in diverse cellular locations. Among these is the regulation of Ca2+-triggered exocytosis at the cell membrane, a process critical to many cellular functions7. In pancreatic β cells, exocytosis is critical for the pulsatile insulin secretion that is regulated by a variety of stimuli, including electric, metabolic and hormonal signals8,9. Ca2+ influx is stimulated when glucose metabolism produces an increase in the [ATP]/[ADP] ratio that inhibits K-ATP channels and induces membrane depolarization, leading to the opening of voltage-gated Ca2+ channels. When activated either by glucose10 via Ca2+ or by hormones such as glucagon-like peptide-1 via classical G-protein-coupled receptor signaling9, PKA can in turn modulate the Ca2+ signal11,12 and also directly influence exocytotic insulin release7. For instance, a PKA-dependent mechanism operates during the initial phase of glucose-induced excytosis13, and impairment of this process has been implicated in the pathogenesis of type 2 diabetes14.
Here, we investigated the role of the temporal regulation of PKA in integrating and transducing diverse signals. By combining fluorescent biosensor-based live-cell tracking of signaling activities with mechanistic modeling, we demonstrate that PKA exhibits oscillatory activity and, together with Ca2+ and cAMP, forms a highly integrated oscillatory circuit in MIN6 β cells. The organization of this oscillatory circuit allows PKA to modulate the frequency of oscillation, to integrate input signals and to exert diverse spatiotemporal controls on substrate phosphorylation. This demonstration of a functional role for the temporal regulation of a protein kinase in achieving signaling diversity helps to establish a new model for encoding signaling information into temporal parameters, such as the frequency of oscillation, of an enzymatic activity.
We set out to analyze the temporal patterns of PKA activity in MIN6 β cells treated with tetraethylammonium chloride (TEA), a pharmacological agent that induces membrane depolarization to trigger a Ca2+ signal15. Under this condition, both Ca2+ and cAMP have been shown to undergo oscillations16. It is therefore interesting to examine the activity pattern of PKA in this context because PKA activity may exhibit a range of dynamic behaviors depending on the relationship between the dynamics of the cAMP oscillatory input and PKA activation/inactivation kinetics. As illustrated in Figure 1A (See Section VIIIA of Supplementary Methods for details), PKA activity could exhibit monotonic increases when the input oscillation frequency is much greater than the rates of the PKA activation/inactivation kinetics (top panel, Figure 1A), oscillatory changes when these values match (bottom panel, Figure 1A), or semi-oscillatory intermediate responses (middle panel, Figure 1A).
We first analyzed PKA activity at the cell population level17,18, using an antibody against the phosphorylated consensus substrate motif of PKA. Sustained monotonic increases in substrate phosphorylation by PKA were observed in MIN6 cells treated with TEA (Supplementary Figure 1), suggesting that PKA activity might not conform to cAMP oscillations or that the oscillatory PKA activity is masked at the cell-population level. To analyze the temporal dynamics of PKA at the single-cell level, we performed fluorescence imaging of endogenous PKA activity using a genetically encoded A-kinase activity reporter (AKAR)19,20. In contrast to the population results, this analysis revealed oscillatory PKA activity (Figure 1B and 1C), suggesting a repetitive shifting of the balance between the actions of PKA and counteracting phosphatases in single cells. These oscillations occurred with a period of 1–6 min in >60% of AKAR-expressing MIN6 cells (n = 21), matching the frequency of the Ca2+ oscillation patterns previously reported in these cells16. Therefore, PKA activity exhibits oscillatory changes in single MIN6 cells, which were likely masked in cell population studies because the observed oscillations in PKA activities were asynchronous, varying in phase and frequency among individual cells.
Given that Ca2+ is known to oscillate under the same conditions and to serve as an important signal in β cells, we characterized the temporal correlation between PKA activity and Ca2+ dynamics by performing co-imaging experiments using AKAR in conjunction with the Ca2+ dye Fura-221 in single MIN6 cells. For these experiments, we engineered a red-shifted AKAR, AKAR-GR (Figure 2A and Supplementary Figure 2A) based on EGFP, and the red fluorescent protein mCherry, to allow better signal separation from Fura-2. We found that the PKA oscillations and Ca2+ oscillations were “in register” with each other (Figure 2B). Notably, during each cycle of oscillation, the rise in the Ca2+ transient appeared to follow, with a noticeable lag, the accumulation of PKA activity (Figure 2B). This pattern likely reflected the PKA-mediated potentiation of Ca2+ influx via voltage-gated Ca2+ channels and/or inositol trisphosphate receptors (IP3R)22,23. Moreover, the decrease in PKA activity in each cycle was surprisingly rapid when compared to the transient PKA activities observed in other cell types (Supplementary Figure 3)24, and it was precisely correlated with the Ca2+ spike, suggesting that there is a Ca2+-dependent mechanism in place to ensure the rapid down-regulation of and/or counterbalancing of PKA activity.
In MIN6 cells, Ca2+-dependent regulation of PKA can occur through Ca2+ dependent adenylate cyclases (ACs) and phosphodiesterases (PDEs)7, the enzymes that produce and degrade cAMP, respectively. Indeed, as previously observed,16 cAMP was found to oscillate under the same conditions when visualized with a fluorescent cAMP indicator, ICUE25,26 (Supplementary Figure 4). Simultaneous imaging using a red-shifted ICUE, ICUE-YR (Figure 2A, and Supplementary Figure 2B), and Fura-2 in single MIN6 cells confirmed that these cAMP oscillations are highly synchronized with the Ca2+ oscillations (Figure 2C).
The high degree of synchronization between PKA and Ca2+ and between cAMP and Ca2+ suggested that all three components were regulated in a temporally coordinated fashion. Since PKA activity is often compartmentalized in various systems3, we asked if the cAMP and PKA oscillations were also correlated in the local cellular context. To answer this question, we took advantage of a new single-chain dual-specificity biosensor for parallel detection of PKA activity and cAMP dynamics (Figure 2A, and Supplementary Figure 5). This biosensor contains both cAMP-sensitive and PKA-dependent molecular switches and can simultaneously capture the dynamics of two signaling events in the same local environment, thereby facilitating quantification and correlation of these signaling activities. In MIN6 cells expressing this single-chain dual-specificity biosensor, TEA treatment elicited oscillatory signals from both reporting units (Figure 2D). These oscillations were highly temporally coordinated, with cAMP rises always preceding increases in PKA activity (Figure 2D), suggesting that oscillations in PKA activities are indeed locally driven by oscillatory changes in cAMP levels.
The precise temporal correlation between the oscillatory activities of Ca2+, cAMP and PKA suggests a complex underlying mechanism involving multiple feedback interactions. To better understand this signaling circuit, we constructed a mathematical model by incorporating a detailed description of the biochemical events involving these three key molecular components. In our model (described in detail in the Supplementary Methods), the interplay between the membrane-potential regulation and [Ca2+] forms the first feedback loop, consistent with the Chay-Keizer model27. A second feedback loop, centered around Ca2+, cAMP and PKA, was incorporated by explicitly introducing putative regulation of both voltage-gated L-type Ca2+ channels and IP3R by active PKA22,23. Additional links needed to complete the model were introduced on the basis of a network topology analysis determining the minimal model sufficient to account for several qualitative features of our experimental findings (see Supplementary Methods Section III for complete model description and analysis). When operating under assumptions about parameter values consistent with literature data, the model was able to capture the detailed dynamic relationships between various molecular components observed in our experiments (Figure 2E-G). Our combined experimental and modeling data therefore suggested the existence of a functional signaling circuit consisting of Ca2+, cAMP and PKA. In this circuit, the activities of all the components are temporally regulated in the form of synchronized oscillations, as a result of a tight coupling between these components through feedback interactions.
To dissect the role of PKA in this oscillatory circuit, we examined the effects of altered PKA activity on the oscillatory patterns of the Ca2+-cAMP-PKA system. The model predicted that the oscillatory behavior of the system, as indicated by the Ca2+ signal, would strongly depend on the level of PKA activity, although oscillation of PKA activity was not required for maintaining Ca2+ oscillations (Supplementary Figure 6). As shown in Figure 3A, inhibition of PKA activity was predicted to have strong abolishing effects on Ca2+ oscillations. Indeed, addition of the PKA inhibitor H89 to MIN6 cells rapidly terminated Ca2+ oscillations, indicating that PKA activity was required (Figure 3B and Supplementary Figure 7A). Consistent with this observation, no TEA-induced Ca2+ oscillations were observed in 17 of 24 cells in the presence of a specific PKA inhibitor, mCherry-tagged PKI. This result represents a 3-fold reduction in oscillatory behavior compared to control cells, over 87% of which showed Ca2+ oscillations (Supplementary Figure 7C). In contrast, the model predicted that increasing PKA activity would modulate the frequency of the Ca2+ oscillations (Figure 3C). This prediction was tested by monitoring Ca2+ dynamics when a general phosphodiesterase inhibitor, 3-isobutyl-1-methylxanthine (IBMX), was added to MIN6 cells after TEA-induced Ca2+ oscillations were established. We found that every cell showing regular oscillations both before and after IBMX addition exhibited an increase in oscillation frequency (n = 15) (Figure 3D and Supplementary Figure 7B), demonstrating that cAMP/PKA activity can modulate the Ca2+ oscillation frequency.
On the other hand, PKA-mediated phosphorylation is continuously counterbalanced by phosphatase activity. For instance, PKA and phosphatase PP2B are known to have opposing effects on L-type Ca2+ channels via phosphorylation and dephosphorylation of the channels, respectively28. Our model predicted that changes in PP2B activity would have an effect opposite to that of PKA activity in terms of modulating the frequency of Ca2+ oscillations (Figure 3E). Consistent with this prediction, treatment of MIN6 cells with 3 μM cyclosporine A, a PP2B inhibitor, increased the frequency of TEA-induced Ca2+ oscillations (n = 7) (Figure 3F), indicating that the ability of PKA to modulate Ca2+ oscillation is under the influence of opposing phosphatase activities.
Interestingly, model simulations also showed that when compared to the Ca2+ oscillation amplitude, the oscillation frequency was considerably more sensitive to changes in PKA activity and variation of other model parameters (Figure 3G, 3H and Supplementary Figures 8–12). This finding suggests a role for PKA as a frequency modulator. Taken together, these data suggest that PKA activity can directly modulate the oscillatory patterns of the Ca2+-cAMP-PKA signaling circuit, thereby affecting cellular processes under the control of Ca2+.
β cells are constantly subject to regulation by a broad range of distinct signals, including electric, metabolic and hormonal signals8,9. Given the tunability of this signaling circuit and the capacity of its major components to connect to different signal-sensing networks29,30, one of its important functions may be to integrate multiple input signals that can independently activate the circuit’s components. Thus, it is possible that various signals, via direct activation of different components, can trigger the oscillations of this signaling system. To test this hypothesis, 1–3 μM of 8-bromoadenosine-3′,5′-cyclic monophosphate-acetoxymethylester, a cAMP analog that preferentially activates PKA, was added to MIN6 cells expressing AKAR. In over 40% of the cells (n = 16), the addition of this cAMP analog alone triggered oscillatory changes in PKA activity (Figure 4A and Supplementary Figure 13), similar to the TEA-induced oscillations. Furthermore, different strengths of the input signals could apparently be encoded into different activity patterns, because higher doses (10–20 μM) of the same cAMP analog triggered a gradual increase in PKA activity with a small-amplitude higher-frequency oscillatory change superimposed on the monotonic activation rise (Figure 4A) (n = 13). Our mathematical model captured the basic characteristics of the oscillations induced by direct activation of PKA (Figure 4B) and suggested that sufficiently strong PKA feedback can initiate oscillations even when the feedback loop involving Ca2+ and membrane potential is initially inactive. Furthermore, the distinct PKA activity patterns induced by low doses and high doses of cAMP (Figure 4) showed an elaborate transition of behaviors that is indicative of the frequency change of the oscillatory circuit. Thus, PKA is an integral component of the oscillatory circuit and is capable of not only tuning but also enabling the oscillatory circuit behaviors in response to specific input signals.
Each of the key components of the oscillatory circuit — Ca2+, cAMP and PKA, regulates a diverse set of cellular functions with high specificity. In the case of PKA, spatial compartmentation of PKA-dependent signaling has been identified as an important mechanism underlying specific regulation of distinct functions3. We propose that the oscillatory PKA activity enabled by the Ca2+-cAMP-PKA circuit can serve as a distinct mechanism for achieving signaling diversity.
At the level of temporal regulation, given an oscillatory PKA activity, both oscillatory and sustained phosphorylation of PKA substrates can be achieved, depending on the access of the substrate to phosphatases reversing PKA phosphorylation (Supplementary Figure 14). For example, recruitment of PKA and phosphatase PP2B to L-type Ca2+ channels can ensure efficient and rapid phosphorylation and dephosphorylation within this signaling complex28. Repetitive phosphorylation of the channels, which has been suggested in previous studies to be critical for proper β cell function31, would in turn ensure that the activity of the channel reaches certain threshold level only periodically; this pattern would decrease the probability of spontaneous channel openings between insulin secretion phases. In contrast, PKA targets devoid of high local phosphatase activity can integrate multiple oscillation phases over time, thus enabling a distinct mode of regulation.
More exquisite spatial control of PKA activity may also be achieved via oscillations. For instance, cAMP/PKA signaling microdomains, such as those defined by AKAP-assembled signaling complexes, may also depend on the kinetics of signaling reactions and the presence of regulatory feedback interactions32. A local PKA signaling complex is typically assembled by A-Kinase anchoring protein (AKAP) binding to the regulatory (R) but not the catalytic (C) subunit of PKA. This signaling compartmentalization can be lost if the unanchored C-subunit diffuses away after dissociation from the R subunit, rather than being recaptured when cAMP dissociates from the R subunits. It is thus paradoxical that local PKA activity can be maintained for prolonged periods of time. Our model suggests that, under the experimental conditions analyzed above, this apparent paradox can be resolved, since the oscillatory circuit is capable of maintaining a relatively low time-averaged PKA activity, while having relatively high maximal peak PKA activity (Figure 5A; local activation regime). The low average PKA activity implies a low average free concentration of the C-subunit in the signaling domain and thus a low escape rate of the subunit from the domain. Recapture of the C-subunit can help reset the local PKA activity cycle. At the same time, the peak PKA activity can periodically exceed the threshold levels needed for full activation of the target substrates and would be further enhanced if the substrates have limited access to the corresponding phosphatases (Figure 5A). On the other hand, an enhanced input into PKA activation and the resulting high average PKA activity could lead to increased escape of the C subunit from a local signaling domain (Figure 5A; global activation regime). At the subcellular organelle level, increased escape of the C-subunit from a cytosolic signaling compartment could result in the C-subunit’s translocation into the nucleus (Supplementary Figure 15). We sought to experimentally validate this prediction. Indeed, we found that while low doses of a cAMP analog induced oscillations in the cytosol without affecting nuclear PKA activity or phosphorylation of an endogenous transcription factor CREB, the gradual increase in PKA activity with superimposed high-frequency oscillation induced by high doses of the analog led to an increase in nuclear PKA activity and CREB phosphorylation (Figure 5B - D and Supplementary Figure 16). This effect on CREB phosphorylation was also observed with the co-treatment of CsA or IBMX with a low dose of the cAMP analog (Supplementary Figure 17) or with TEA (Supplementary Figure 18). Thus, the oscillatory PKA activity can apparently encode context-dependent spatiotemporal information for controlling diverse output signals.
Oscillation of signaling activity can be a versatile mechanism for achieving diversity and specificity. This notion was first demonstrated in the case of Ca2+ oscillation, the frequencies of which are determined by the amount and type of extracellular signals33–35 and can influence the efficiency and specificity of gene expression36,37. Our analysis suggests that, as in the case of Ca2+ oscillations, the frequency of PKA activity oscillation can be used as a signaling currency, integrating input signals according to their identities and doses and generating specific functional output (e.g., local vs. global signaling modes). In this way, the oscillatory PKA activity allows PKA to function as a versatile signaling transducer that can exert spatiotemporal controls over a variety of functional outcomes in response to specific signals. Future studies will directly evaluate the effects of changing the PKA oscillation frequency on cellular processes that are important to β-cell functions, such as gene expression and insulin secretion.
Our analysis demonstrates that, in addition to transducing signals via activity oscillation, PKA in insulin-secreting cells also plays an active role in shaping the oscillatory patterns of other key components in the signaling circuit, such as Ca2+. This finding is in contrast to previous studies in which oscillations of several protein kinases (e.g., PKA38, PKC39, mitogen-activated protein kinases18,40, and calmodulin-dependent protein kinase II41) were demonstrated, but the potential role of these kinases in shaping the oscillation patterns was not explored. It remains to be seen whether some of these kinases or other candidate kinases are also focal points of oscillatory activities.
Given that each of the key components in the Ca2+-cAMP-PKA circuit can regulate distinct functions, the synchronized oscillations ensure the coordinated activity of these key players, providing a mechanism for achieving efficient and specific control of cellular processes. Therefore, it is reasonable to suggest that complex cellular behaviors may critically depend on such synchronization and coordination of various signaling molecules.
In summary, this current study provides a clear example of exquisite temporal control of a protein kinase within an oscillatory circuit. This signaling system could serve as a paradigm for understanding the spatiotemporal regulation of other signaling networks that control complex cellular processes. Furthermore, given the ubiquitous presence of the key components of the oscillatory Ca2+-cAMP-PKA circuit42,43, its effects are likely to be widespread, serving to regulate diverse processes in various cell types.
EGFP and mCherry were PCR amplified to introduce appropriate restriction digest sites for subcloning into either AKAR or ICUE constructs. The genes of GR-AKAR and YR-ICUE were generated in pRSETB (Invitrogen) and then moved to pcDNA3 (Invitrogen) behind a Kozak sequence for mammalian expression. ICUEPID was generated from CRY-AKAR44 in pcDNA3 by replacing mVenus with the molecular switch and mCitrine of ICUE2.
MIN6 cells were plated onto sterilized glass coverslips in 35-mm dishes and grown to 50–90% confluency in DMEM (10% FBS) at 37°C with 5% CO2. Cells were transfected using Lipofactamine 2000 (Invitrogen) and grown 20–48 h before imaging. For Ca2+ imaging experiments, cells were preincubated with 1 μM Fura-2/AM (Molecular Probes) for 10–20 min at 37 °C before imaging
Cells were washed twice with Hanks’ balanced salt solution buffer and maintained in the dark at room temperature. Cells were imaged on a Zeiss Axiovert 200M microscope with a cooled charge-coupled device camera (MicroMAX BFT512, Roper Scientific, Trenton, NJ) controlled by METAFLUOR 6.2 software (Universal Imaging, Downingtown, PA). Fura-2 dual excitation ratio imaging used two excitation filters [350DF10 for 350 nm excitation and 380DF10 for 380 nm excitation], a 450DRLP dichroic mirror, and a 535DF45 emission filter. Dual red/green emission ratio imaging used a 480DF30 excitation filter, a 505DRLP dichroic mirror, and two emission filters [535DF45 for GFP and 653DF95 for RFP]. Dual red/cyan emission ratio imaging used a 420DF20 excitation filter, a 450DRLP dichroic mirror, and two emission filters [475DF40 for CFP and 653DF95 for RFP]. Dual yellow/red emission ratio imaging used a 495DF10 excitation filter, a 515DRLP dichroic mirror, and two emission filters [535DF25 for YFP and 653DF95 for RFP]. Dual cyan/yellow emission ratio imaging used a 420DF20 excitation filter, a 450DRLP dichroic mirror, and two emission filters [475DF40 for CFP and 535DF25 for YFP]. These filters were alternated by a filter-changer Lambda 10–2(Sutter Instruments, Novato, CA). Exposure time was 50–500 ms, and images were taken every 10–30 s. Fluorescence images were background-corrected by subtracting the fluorescence intensity of background with no cells from the emission intensities of cells expressing fluorescent reporters. The ratios of 350/380, red/green, red/cyan, yellow/red, or yellow/cyan emissions were then calculated at different time points. The values of all time courses were normalized by dividing each by the average basal value before drug addition.
Simulations were carried out with custom written MATLAB programs. The system has 6 states: membrane potential (V), gating probability of K+ channel (w), concentration of cytosolic Ca2+ ([Ca2+]i), [cAMP], concentration of cAMP bound regulatory subunits of PKA ([c4R2]), concentration of free catalytic subunit which is the active PKA ([PKA*]). Kinetic equations for Ca2+ flux through the plasma membrane were based on the Chay-Keizer model and the Ca2+ release from internal stores was modeled as arising out of the coordinated activities of IP3Rs and SERCAs. Both the fluxes were dependent on PKA activity, thereby setting up a feedback loop. The increase in [Ca2+]i modulates the activities of Ca2+ inhibitable AC VI and Ca2+ activated PDE1, which in turn produce and degrade cAMP respectively. PKA is activated by the sequential binding of 4 molecules of cAMP and the catalytic subunit is released sequentially. Concentrations were normalized with respect to their respective maximal levels, for the ease of comparison. The programs were simulated for a sufficient time span (1000 s for traces and 3000 s for parameter tuning analysis) to produce sustained oscillatory patterns. Frequency of Ca2+ oscillations was calculated using MATLAB’s in-built FFT routine. A detailed description of the model and the derivation of the ODEs are given in the Supplementary Methods.
For analysis of oscillatory changes in PKA activity, MIN6 cells were stimulated with 20 mM TEA and washed with ice-cold phosphate-buffered saline at different time points (0, 1, 2, 3, 4, 5, 6, 7, 8 min). Cells were rapidly lysed in radioimmunoprecipitation assay lysis buffer containing protease inhibitor cocktail, 1 mM ethylenediaminetetraacetic acid (EDTA), 1 mM phenylmethylsulfonyl fluoride(PMSF), 1 mM NaVO4, 1 mM NaF, and 25 nM Calyculin A. Cell lysates were incubated on ice for 30 min and centrifuged at 4°C for 20 min. Total protein was separated with 7.5 % SDS-polyacrylamide gel electrophoresis (PAGE) and transferred to nitrocellulose membranes, which were probed with phospho-PKA substrate or β-tubulin antibody. Bands were visualized by enhanced chemiluminescence. The intensity of the bands was quantified, and the values were then normalized to β-tubulin level by using UN-SCAN-IT (Silk Scientific, Orem, UT). For detection of nuclear PKA activity, MIN6 cells were stimulated with low dose (2 μM) or high dose (10 μM) of cAMP analog, or 50 μM forskolin (FSK) and washed with ice-cold phosphate-buffered saline after 45 min incubation. Cells were then lysed as described above. Total protein was separated with 7.5 % SDS-PAGE and transferred to nitrocellulose membranes, then probed with phospho-CREB (pS133) or CREB antibody. The intensity of the bands was quantified. The values of phospho-CREB level then were normalized to CREB expression level.
We thank Drs. Jun-Ichi Miyazaki, George G. Holz and Gerald H. Hart for providing cell lines. We also thank Xiong Li for initial technical assistance. This work was supported by National Institutes of Health grants R01 DK073368 and DP1OD006419 (to J. Z.); and GM072024 and RR020839 (to A.L.).
Author ContributionsQ.N. and J. Z. conceived the experimental aspect of the project and did the initial experiments; A. L. designed the modeling aspect of the project. Q.N. and N.A. performed the majority of the experiments. A.G. constructed the mathematic model and performed the simulations. G. X. performed the western analyses. M.D.A. designed and generated one of the biosensors. J. Z., Q. N. and A.L. wrote the manuscript.
Competing Financial Interests
The authors declare no competing financial interests.