Nitric oxide (NO) is a key signaling molecule that regulates diverse biological processes in both animals and plants, including important roles in male gamete physiology. In plants, NO is generated in pollen tubes (PTs) and affects intracellular responses through the modulation of Ca2+ signaling, actin organization, vesicle trafficking and cell wall deposition, bearing consequences in pollen-stigma interactions and PT guidance. In contrast, the NO-responsive proteins that mediate these responses remain elusive. Here, we show that PTs of Arabidopsis thaliana mutants impaired in the pollen-specific DIACYLGLYCEROL KINASE4 (DGK4) grow slower and become partially insensitive to NO-dependent growth inhibition and re-orientation responses. Recombinant DGK4 protein yields NO-responsive spectral and catalytic changes in vitro that are compatible with a role in NO perception and signaling in PTs. In addition to the expected phosphatidic acid-producing kinase activity, DGK4 recombinant protein also revealed guanylyl cyclase activity, as inferred by sequence analysis. Our results are compatible with a role for the fast-diffusible NO gas in signaling and cell-cell communication via the modulation of DGK4 activity during the progamic phase of angiosperm reproduction.
Nitric oxide (NO) is a key signaling molecule that regulates diverse biological processes in animals and plants (Wendehenne et al., 2001; Lamattina et al., 2003). In animals, NO regulates vascular wall tone, neurotransmission and immune response, whereas in plants, NO is essential for development and for responses to biotic and abiotic stresses (Wendehenne et al., 2001; Domingos et al., 2015; Astier et al., 2019). Interestingly, NO is involved in the sexual reproduction of both animals and plants, mediating events related to the male gamete physiology. In animals, NO stimulates sperm motility (Miraglia et al., 2011) and binding to the plasma membrane of oocytes (Sengoku et al., 1998), whereas in plants, NO mediates pollen-stigma interactions and pollen tube (PT) guidance (McInnis et al., 2006; Prado et al., 2008). NO generation in PTs has been demonstrated (Prado et al., 2004) and intracellular responses to NO include modulation of cytosolic Ca2+ elevation (likely causal for PT growth retardation, re-orientation and re-growth), actin organization, vesicle trafficking and cell wall deposition (Prado et al., 2008; Wang et al., 2009). However, the NO-responsive proteins that mediate these responses remain unknown (Domingos et al., 2015; León and Costa-Broseta, 2019).
Previously, pollen-specific Arabidopsis DIACYLGLYCEROL KINASE4 (DGK4; At5g57690) has been suggested to harbor a gas-sensing region (Wong et al., 2013; Domingos et al., 2015), and the DGK family has been associated with important roles in pollen germination and growth (Potocký et al., 2014). Recently, DGK4 has been assigned signaling roles related to maintaining mechanical properties in PT growth with consequences on fertility rates (Vaz Dias et al., 2019).
Here, we have investigated the NO-dependent PT growth responses in wild-type (WT) (ecotype Col-0) and DGK4-impaired plants. We show that PTs of Arabidopsis impaired in the pollen-specific DGK4, grow slower and become partially insensitive to NO-dependent growth inhibition and re-orientation responses. Recombinant DGK4 yields NO-responsive spectral and catalytic changes in vitro that are compatible with a NO perception and signaling function in PTs. These results are discussed in the context of a role for NO, as a fast-diffusible gas, in DGK4-mediated long-range signaling and/or rapid cell-cell communication during angiosperm reproduction.
RESULTS AND DISCUSSION
DGK4 is required for NO-dependent PT growth and re-orientation responses
Homozygous dgk4 plants with a T-DNA insertion 424 bp upstream of the DGK4 open reading frame (dgk4-1; SALK_151239) were determined to have a 50% knockdown (KD) in DGK4 expression (Fig. S1). The PTs of dgk4-1 grow significantly slower in vitro compared with WT across a pH range of 6.5 to 8.5, and in both optimal and reduced Ca2+ media (Fig. 1A). Four hours after germination, WT PTs reached an average length of 176 µm, whereas dgk4-1 only reached 144 µm at pH 7.5 (n>100; P<0.05). This mutant line was also reported to have PTs with altered stiffness and adhesion properties (Vaz Dias et al., 2019). When exposed to the NO donor sodium nitroprusside (SNP), both WT and the dgk4-1 PTs show NO dose-dependent reduction of growth rates (Fig. 1B), much like those reported in Lilium longiflorum (Prado et al., 2004), Paulownia tomentosa (He et al., 2007) and Camelia sinensis (Wang et al., 2012). At doses of between 0 and 10 nM of SNP, the decline in growth rate for WT is 34.52% but, in contrast, the decline for dgk4-1 in the same range is only 7.86%. Importantly, PTs of dgk4-1 become insensitive to SNP concentration increases over 50 nM, whereas the rate of WT PT growth continues to decrease up to 200 nM SNP (Fig. 1B,C). Correspondingly, we also observed differential sensitivity between WT and dgk4-1 PTs using another NO donor, diethylamine (DEA) NONOate (Fig. 1B). A second independent homozygous dgk4 mutant plant (dgk4-2; SALK_145081, T-DNA insertion 268 bp upstream of the DGK4 gene) (Fig. S1) showed similar NO insensitivity (Fig. S2). These results suggest that the effect of NO on DGK4 activity has mechanistic consequences on PT growth.
We further examined the effect of NO on the directional growth of PTs using dgk4-1. Our experimental set-up was based on our previous work in lily PTs (Prado et al., 2004) in which an agarified medium containing 10 mM SNAP (s-nitroso-acethilpenilcilamine; NO donor) was loaded into a pulled glass micropipette, forming a point source of NO diffusion. Assuming that the germination medium is homogenous, a gradient concentration decrease proportional to the square of the distance from the source should be formed. For lily PTs, we experimentally determined that the threshold for the negative chemotropic re-orientation response was within the range of 5-10 nM of NO (Prado et al., 2004). Here, we reproduced these experiments with Arabidopsis. Both WT and dgk-4 PTs were chosen to have growth rates within a similar range (2-4 µm min−1) to minimize growth rate bias to their NO response. We observed that WT and dgk4-1 PTs showed a negative chemotropic response, bending away from the NO source, much like our previous observations in lily (Prado et al., 2004); however, of relevance, the bending angles in WT (45.8±6.3°; n=10) were twice as acute as those in dgk4-1 (22.9±2.6°; n=10) (Fig. 1D; Fig. S3). Moreover, WT PTs showed clear, elbow-like points of inflections when approaching the critical NO concentration, a morphological feature also observed in lily (Prado et al., 2004) that corresponds to a slowdown of the growth rate, followed by turning of the PT (Movies 1,2). In contrast, dgk4-1 PTs showed softer angles without elbow-like turns (Fig. 1D), even when challenged with much steeper and concentrated gradients of SNAP (Fig. S3). Both the germination and the re-orientation experiments clearly show a desensitization to NO, an effect that would be expected to be more pronounced if a full knockout of dgk4-1 was available. Given that a dose of 5-10 nM is within the physiological range of NO action (Lamattina et al., 2003), our results are consistent with a signaling role for DGK4 in NO sensing.
In agreement with the in vitro germination phenotype, in vivo germinated PTs of dgk4-1 also grew slower down the pistil than those of WT across all time points examined (Fig. 1E,F). To test whether these differential growth rates would result in outcompetition, we brushed a ∼50:50 mixture of pollen from Col-0 (rr) and dgk4-1 (RR) onto an emasculated Col-0 flower (where R=resistant and r=susceptible), generating a Rr pollen cross onto a Col-0 female (rr). With no bias in fitness, a 1:1 ratio of kanamycin-resistant and susceptible offspring would be expected, but rather we observed a percentage of 36.3% (n=99) kanamycin-resistant offspring, revealing that WT PTs outperformed dgk4-1 PTs. This finding is consistent with that reported for a different dgk4 mutant allele (Vaz Dias et al., 2019).
DGK4 yields NO-responsive spectral changes and catalytic activity
We next focused on determining the molecular basis of the NO-sensing properties of DGK4. Through sequence analysis we have previously predicted that DGK4 contains a region spanning from H350 to R383, similar to heme centers of functional gas-responsive heme-NO/oxygen (H-NOX), heme-NO-binding and NO-sensing families of proteins in other kingdoms (Domingos et al., 2015; Wong et al., 2018). In particular, this region harbors the HXPX[14,16]YXSXR consensus pattern derived from heme b containing H-NOX centers in proteins from bacteria and animals, and is present in plant orthologs, such as poplar, castor bean and soybean, but absent in other Arabidopsis DGKs (Fig. 2A). The presence of the H-NOX-like signature suggests that DGK4 might accommodate a heme b and, correspondingly, the diagnostic spectral properties should have a distinct response to NO. Recombinant DGK4 yields a Soret peak at 410 nm (Fig. 2B), which is distinctly different from unbound hemin (protohemine IX: Soret band at 435 nm with a shoulder at 400 nm) and falls within the typical peak range observed for proteins with a histidine-ligated ferric heme b (Walker et al., 1999). Reduction with sodium dithionite resulted in a red shift of the Soret peak to 424 nm accompanied by the emergence of distinct α (558 nm) and β (526 nm) bands (Fig. 2B; Fig. S4A). The ferrous state presumably represents the native state of DGK4 in the cytosol [Arabidopsis cytosolic redox potential: −310 to −240 mV (Aller et al., 2013)]. Exposure to air recovers the oxidized Soret peak (410 nm) of DGK4 after 20 min (Fig. S4A). Importantly, addition of DEA NONOate attenuates the reduced Soret absorption (424 nm) in a concentration-dependent manner, hinting at the possibility of NO displacing the histidine ligand from the heme group (Fig. 2C; Fig. S4B). This is an essential step in the signaling of canonical H-NOX proteins (Russwurm and Koesling, 2004).
Qualitatively, the spectroscopic behavior resembles that of canonical H-NOX proteins, e.g. the H-NOX domain of Shewanella oneidensis, which showed Soret absorptions at 403 nm (ferric), 430 nm (ferrous) and 399 nm (ferrous, NO bound) (Dai et al., 2012). However, the frequencies and relative intensities of the Soret α and β peaks are indicative of a bis-histidine ligated heme b center, as found in cytochromes b5 or a heme-based cis-trans carotene isomerase Z-ISO (Beltrán et al., 2015). In accordance, DGK4 mutations, which affect the heme binding site, should produce a reduced heme absorption spectrum, as demonstrated for Z-ISO (Beltrán et al., 2015). Indeed, H350L and Y379L dgk4 mutants showed reduced Soret band intensities of about 50% and 70%, respectively (Fig. 2D). As the Soret bands were still present in the mutants, albeit attenuated, we can expect a similar behavior in their reductions and NO spectra, which we did observe with the H350L mutant protein (Fig. S4). Overall, the H350L dgk4 mutant recorded a much larger decrease in reduced Soret bands than observed with DGK4 WT at low NO donor concentration (0.25 mM DEA NONOate) and also required a slightly longer time (∼5 min more than DGK4 WT) to recover its oxidized Soret peak (410 nm) when exposed to air (Fig. S4). Together with a marked reduction in Soret band intensities of dgk4 mutants, these results can be interpreted as a weakening of the heme environment. Our data are consistent with a recently identified Arabidopsis protein, AtLRB3 (At4g01160), which, like DGK4, harbors a H-NOX-like center that yields canonical spectroscopic behavior of H-NOX proteins, and for which mutation of the heme iron-coordinating histidine impairs the NO response (Zarban et al., 2019).
Based on protein sequence analysis we previously predicted DGK4 to be a bifunctional catalytic protein with a canonical kinase domain capable of converting sn-1,2-diacylglycerol (DAG) with ATP into the corresponding phosphatidic acid (PA), but, importantly, with a hypothetical guanylyl cyclase (GC) center capable of catalyzing the formation of cGMP from GTP (Wong et al., 2013; Domingos et al., 2015; Xu et al., 2018; Su et al., 2019). Our prediction was recently confirmed by others (Vaz Dias et al., 2019). Thus, we next focused on the catalytic activity of DGK4. Both cGMP and, more strongly, NO inhibited DGK4 kinase activity but NO did not affect the GC activity (Fig. 2E). Mutations in the H-NOX center did not affect the kinase activity of DGK4, as both H350L and Y379L dgk4 mutants were functional and inhibited by NO to a similar degree as the WT (Fig. S5). Although this result could nullify the hypothesis that the H-NOX center of dgk4-1 works as a NO sensor, it must be considered that, although there is a reduction of intensity and slower recovery upon oxidation, there are still changes in the Soret band in the mutants, thus revealing some NO binding. Given that both the enzyme and the NO steady concentrations in vivo should be much lower (Prado et al., 2004), the range of physiological sensitization to NO might therefore not be captured given the conditions of the enzymatic assays.
Another conceivable possibility that we cannot overrule at this point is that NO-induced S-nitrosylation is also taking place, as reported for other kinases (Hu et al., 2017; Liu et al., 2017). DGK4 displays five S-nitrosylation sites on C62, C138, C174, C260 and C334 (SNOSite, csb.cse.yzu.edu.tw/SNOSite/Prediction.html; Lee et al., 2011). Of these, C138 and C174 are in the kinase/GC domains but no S-nitrosylation sites have been predicted at the H-NOX center. Either way, as DGK4 is localized in the cytosolic region of the PT apex (Vaz Dias et al., 2019), our results are consistent with a signaling role for DGK4 in transducing NO-binding into lipid-, cGMP- or Ca2+-dependent downstream responses. This conclusion is based on the fact that DGKs are known to convert DAG to PA, which in turn is essential for PT growth (Pleskot et al., 2012; Potocký et al., 2014) and the mobilization of Ca2+ (Monteiro et al., 2005; Prado et al., 2008). Our biochemical data also show that, although the kinase activity of DGK4 is inhibited by cGMP, the GC activity of DGK4 is unaffected by NO, thus implying that the reduced NO response of dgk4-1 PTs might be achieved primarily through the PA-dependent lipid/Ca2+ signaling pathways rather than the activation of its GC moonlighting center. Although the existence of a cyclic nucleotide signaling paradigm in plants is controversial (Gehring and Turek, 2017; Świeżawska et al., 2018), cGMP has been shown to activate Ca2+ currents by the cyclic nucleotide-gated channel 18 (CNGC18) localized at the tip of PTs (Gao et al., 2016).
In summary, Fig. 2F depicts a hypothetical mechanism for NO action in PTs. In this model, NO could have two distinct targets in DGK4: on the one hand, NO modulates PT re-orientation responses by binding to the H-NOX center of DGK4, leading to a PA reduction; on the other hand, NO inhibits the kinase and GC activities of DGK4 via S-nitrosylation at C138 and/or C174. We hypothesize that differential kinetics of these mechanisms (H-NOX is predicted to occur much faster), different affinities of NO binding (defining different intervals of NO perception), and the role of cGMP (which we found to inhibit the kinase activity but is not affected by NO) could generate the sequential effects observed during our slowing/re-orientation assays, and the lack of thereof in the mutant. The readout of DGK-4-mediated reactions would be transduced into variations of PA, which is known to interplay with Ca2+ signaling, possibly triggering other downstream paths resulting in vesicular trafficking and/or alterations in actin dynamics (Li et al., 1999; Chen et al., 2003). In agreement with this interpretation, dgk4 PTs have recently been reported to exhibit altered mechanical properties with downregulation of a cyclase-associated protein CAP1, which is involved in actin dynamics, and a general downregulation of phosphoinositide metabolism (Vaz Dias et al., 2019).
Importantly, the slowing down of PTs is necessary for optimizing the perception of chemical cues (Stewman et al., 2010); this constitutes a plausible explanation for the loss of chemotropic response in the mutant and contributes to the observed seed-set reduction. As a fast-diffusible gas, NO is well suited to perform fine-tuning of rapid cell-cell communications, such as the pollen-stigma interactions (Prado et al., 2008; Feijó, 2010; Domingos et al., 2015), and we posit that DGK4 might be an important sensing input mechanism of NO to regulate plant reproduction.
MATERIALS AND METHODS
Plant materials and growth conditions
Two mutant Arabidopsis lines (SALK_151239 and SALK_145081) with T-DNA insertions at the promoter of DGK4 were obtained from the Nottingham Arabidopsis Stock Center and progenies were PCR screened for homozygosity of mutant (T-DNA+DGK4 reverse primer) and WT (DGK4 promoter forward+DGK4 reverse primer) chromosomes (Table S2). The homozygous mutant alleles were subsequently referred to as dgk4-1 and dgk4-2, respectively. The T-DNA insertion sites were confirmed by sequencing (KAUST Bioscience Core Lab). All seeds were stratified at 4°C for 3 days. Col-0 (wild-type) and mutant lines were grown on soil (Jiffy) containing 50% (w/v) of vermiculite in Percival growth chambers (CLF Plant Climatics) at 22±2°C and 60% of relative humidity under long day (16 h light) photoperiod (100 µM photons m−2 s−1).
Characterization of dgk4 mutant plants
RNA was extracted from the pollen of ∼300 flowers of WT and dgk4-1 mutants (Qiagen) and cDNA synthesized using SuperScript III reverse transcriptase according to manufacturer's protocols (Invitrogen). The cDNA was subjected to semi-quantitative RT-PCR with DGK4 gene-specific primers (Table S2) on an AB thermal cycler (Bio-Rad); DGK4 gene expression was normalized against that of protein phosphatase 2A subunit A3, PP2AA3 (At1g13320) (Table S2), using the ImageLab software (Bio-Rad).
In vitro pollen germination
In vitro pollen germination was performed as previously described (Prado et al., 2004, 2008) but with optimized germination medium for Arabidopsis (Boavida and McCormick, 2007) with or without the NO sources sodium nitroprusside (SNP) or DEA NONOate. The growth of at least 100 PTs was measured on a Nikon Eclipse TE2000-S inverted microscope equipped with an Andor iXon3 camera across a range of pH values (pH 6.5-8.5) and in optimal (5 mM) and low (100 µM) Ca2+ media. In NO-treated pollen germination and PT growth experiments, at least 150 unique pollen or PTs were considered. Image frames covering the entire growth area of the culture dish that is mounted on automated stage were acquired using the Nikon Eclipse TE2000-S inverted microscope, which is equipped with a Hamamatsu Flash28s CMOS camera. The PT lengths were measured using NeuronJ (Meijering et al., 2004). For PT re-orientation studies, PTs were challenged with a NO artificial point source by loading with glass micropipettes (OD 1.5 mm, ID 1.2 mm, WPI) pulled to a tip with a ∼5 µm aperture filled with 10 mM SNAP (s-nitroso-acethilpenilcilamine) on agarified (1%) germination medium. Typically, the pipette tip was placed 60 µm away from the growing PT tip using a nanometer-stepper, motor-driven, three-dimensional positioner (Science Wares). Growth and bending responses of growing PTs were monitored by imaging using a Nikon Eclipse TE2000-S inverted microscope (PlanApo 40×, NA 1.3) equipped with an Andor iXon3 camera and the bending angles measured using ImageJ (Schneider et al., 2012).
PT growth in planta
PT growth in the pistil of hand-pollinated WT and dgk4-1 plants was examined by collecting the pistils at different time points (3-8 h) after pollination. Aniline blue staining of PTs in the pistil was performed as described previously (Mori et al., 2006) and PT length in the pistil was measured using ImageJ (Schneider et al., 2012).
Reproductive fitness test
The reproductive fitness of dgk4-1 was examined by crossing emasculated WT (ecotype Col-0) flowers with pollen from both WT and dgk4-1. In detail, a ∼50:50 mixture of pollen from Col-0 (rr) and dgk4-1 (RR) was brushed onto an emasculated Col-0 flower, where pollen was effectively Rr crossed onto the Col-0 female (rr). If there was no bias in fitness, then there should be a 50% kanamycin-resistant and 50% kanamycin-susceptible phenotype. If the Col-0 pollen is fitter, then fewer than half of the offspring will be kanamycin resistant. The desiccated seeds were collected, surface sterilized and stored at 4°C for 3 days and then grown on Murashige and Skoog agar medium (1.1% w/v) containing 100 µg/ml kanamycin (Sigma-Aldrich). The proportion of WT to dgk4-1 seeds were scored after 7 days of growth (n>70).
Protein expression and purification
A Gateway-compatible clone (DKLAT5G57690.1) containing the full-length coding sequence of DGK4 was purchased from Arabidopsis Biological Resource Center. The DGK4 sequence was recombined into the pDEST17 his-tagged expression vector and transformed into E. coli BL21 A1 (Invitrogen). Expression of recombinant DGK4 was induced with 0.2% (w/v) L-arabinose. Cells were lysed in a guanidium lysis buffer and the supernatant loaded onto a Ni-NTA agarose column for affinity purification under denaturing conditions using urea-containing buffers. Denatured recombinant DGK4 was refolded by gradual dilution of urea in a linear gradient using an AKTA fast protein liquid chromatography (GE Healthcare). Hemin (30 µg/ml) was added to the refolding buffers to allow for the incorporation of heme into DGK4 as it assumes native conformation. Excess hemin was removed by size exclusion and recombinant DGK4 stored in ‘buffer’ containing 20 mM Na2H2PO4, 500 mM NaCl, 500 mM sucrose, 100 mM non-detergent sulfobetaines (NDSB), 0.05% (w/v) polyethylene glycol (PEG), 4 mM reduced glutathione, 0.04 mM oxidized glutathione and SIGMAFAST protease inhibitor cocktail (1 tablet per 100 ml solution).
Two single dgk4 mutants (H350L and Y379L) were constructed using site directed mutagenesis by PCR (Ho et al., 1989). To construct the H350L dgk4 mutant, two overlapping fragments of the DGK4 coding sequence both incorporating the mutation, were amplified from the pDEST17-DGK4 plasmid using the respective DGK4 F and DGK4 H-L R (first fragment amplification), and DGK4 H-L F and DGK4 R (second fragment amplification) primer pairs (Table S2). The two overlapping fragments both incorporating the mutations were then used as templates for a PCR reaction using the full-length DGK4 F and DGK4 R primer pairs (Table S2), which generated a full-length dgk4 H350L mutant sequence. The dgk4 Y379L mutant was generated using the same method but with the following mutagenic primers pairs, DGK4 F and DGK4 Y-L R, and DGK4 Y-L F and DGK4 R (Table S2). The DGK4 mutant PCR products were inserted into the PCR8/GW/TOPO vector (Invitrogen) by TA cloning, recombined into the pDEST17 his-tagged expression vector and transformed into E. coli BL21 A1 (Invitrogen). Mutant dgk4 was expressed and affinity purified in the same manner as DGK4.
UV-visible absorption spectroscopy
The UV-visible spectra of affinity purified recombinant DGK4 (200 µg/ml) was recorded on a PHERAstar FS micro-plate reader (BMG Labtech). The heme environment of DGK4 was characterized by the addition of a reducing agent, sodium dithionite (Na2S2O4), to a final concentration of 10 mM and absorbance was immediately measured and examined for spectral changes. The protein sample was then exposed to air and any recovery of the oxidized peak was monitored by the same spectra measurements at 5 min intervals. The heme-NO complex was generated by immediately adding the NO donor DEA NONOate to a pre-reduced recombinant DGK4 before making the same spectral measurements.
DAG kinase and GC assays
DAG kinase assay and phospholipid extraction was performed using 30 μg purified recombinant protein in a reaction mixture containing 40 mM Bis-Tris (pH 7.5), 5 mM MgCl2, 0.1 mM EDTA, 1 mM spermine, 0.5 mM dithiothreitol, 1 mM sodium deoxycholate, 0.02% (v/v) Triton X-100, 500 μM 1,2-DOG and 1 mM ATP in the absence or presence of 1 mM SNP or 0.65 mM DEA NONOate. PA generated from the reactions was measured using the Total Phosphatidic Acid Assay Kit (Cayman Chemical) according to the manufacturer's protocol.
A GC assay was performed using 10 µg purified recombinant protein in a reaction mixture containing 50 mM Tris-HCl (pH 7.5), 1 mM GTP, 5 mM MgCl2 or MnCl2, in the absence or presence of 1 mM SNP or 0.65 mM DEA NONOate. cGMP generated from the reactions was measured using the cGMP enzyme immunoassay (EIA) Biotrak System according to the manufacturer's acetylation protocol (GE Healthcare).
Chemicals and statistical analysis
All chemicals were purchased from Sigma-Aldrich unless stated otherwise. Statistical analyses were performed using an unpaired, one-tailed Student's t-test. Significance was set to a threshold of P<0.05 and n values represent the number of biological replicates.
Conceptualization: A.W., J.A.F., C.G.; Methodology: M.P., J.A.F., C.G.; Validation: M.P., J.A.F.; Formal analysis: A.W., M.P., J.A.F., C.G.; Investigation: A.W., L.D., M.P., J.A.F., C.G.; Data curation: A.W., M.P., J.E., J.A.F., C.G.; Writing - original draft: A.W.; Writing - review & editing: L.D., M.P., J.E., J.A.F., C.G.; Visualization: J.A.F.; Supervision: J.A.F., C.G.; Project administration: J.A.F., C.G.; Funding acquisition: J.A.F., C.G.
This research was supported by the King Abdullah University of Science and Technology (to C.G.); the National Science Foundation (MCB-1616437 and MCB-1930165 to J.A.F.); the National Natural Science Foundation of China (31850410470 to A.W.); and the Zhejiang Provincial Natural Science Foundation of China (LQ19C130001 to A.W.).
Peer review history
The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.183715.reviewer-comments.pdf
The authors declare no competing or financial interests.