Centriole translocation and degeneration during ciliogenesis in Caenorhabditis elegans neurons

Abstract

Neuronal cilia that are formed at the dendritic endings of sensory neurons are essential for sensory perception. However, it remains unclear how the centriole‐derived basal body is positioned to form a template for cilium formation. Using fluorescence time‐lapse microscopy, we show that the centriole translocates from the cell body to the dendrite tip in the Caenorhabditis elegans sensory neurons. The centriolar protein SAS‐5 interacts with the dynein light‐chain LC8 and conditional mutations of cytoplasmic dynein‐1 block centriole translocation and ciliogenesis. The components of the central tube are essential for the biogenesis of centrioles, which later drive ciliogenesis in the dendrite; however, the centriole loses these components at the late stage of centriole translocation and subsequently recruits transition zone and intraflagellar transport proteins. Together, our results provide a comprehensive model of ciliogenesis in sensory neurons and reveal the importance of the dynein‐dependent centriole translocation in this process.

Introduction

Our perceptions of the natural world originate from our senses of an unending variety of stimuli in the environment. Sensory organs have developed specialized structures to perform this feat. For example, the mammalian olfactory system is optimized for the detection and discrimination of odorant molecules through odorant receptor proteins that are expressed by peripheral olfactory sensory neurons (OSNs; McEwen et al, 2008; Challis et al, 2015). A bipolar OSN possesses an unbranched axon that projects to the olfactory bulb and a dendrite that extends apically and terminates in a dendritic knob at the epithelial surface. A remarkable feature of OSNs is the formation of olfactory cilia that emanate from the dendritic knob and project into the mucus of the nasal cavity (Menco, 1997; Jenkins et al, 2009). These cilia harbor G‐protein‐coupled receptors and other proteins that are essential for olfactory signal perception and transduction, making them the sites where the detection of a chemical odor is converted into an electrical response (Berbari et al, 2008). Importantly, each dendritic knob contains 10–30 olfactory cilia of up to 50–60 μm in length, which increases the sensory surface of the ciliated epithelium to amplify the ability of odorant detection (Cuschieri & Bannister, 1975a; Mashukova et al, 2006; Oberland & Neuhaus, 2014). Disruption of ciliary structure or function causes human ciliopathies such as anosmia (inability to perceive odors) or dysosmia (altered odor perception), highlighting the significance of olfactory cilia in odor perception (Toller, 1999; Klysik, 2008; Brown & Witman, 2014).

Cilia are microtubule‐based organelles in which the ninefold symmetric array of doublet axonemal microtubules is extended from the centriole‐derived basal body (Ward et al, 1975; Pedersen et al, 2008; Reiter et al, 2012). The assembly of olfactory cilia has unique features, as multiple cilia are present at the dendritic ending of an OSN (Menco, 1980; Challis et al, 2015; Falk et al, 2015). Cells normally use one centriole to build a single cilium. The multiciliated OSNs are thought to multiply their centrioles in the cell body, and then the centrioles undergo long distance dendritic transport to arrive at the dendritic knob, where they permit the elongation of cilia (Dirksen, 1974; Schwarzenbacher et al, 2005). Immunohistochemistry on ciliogenesis in OSNs has outlined the possible cellular events involved, including dendrite outgrowth, centriole amplification and translocation, and cilium formation (Fig EV1A; Hagiwara et al, 2004; Jenkins et al, 2009). However, the dynamic behavior of centriole and the underlying molecular regulation remain largely unexplored due to the complex developmental processes (at approximately E10–E14 in mouse) and the lack of an experimental methodology for live cell microscopy (Cuschieri & Bannister, 1975a,b; Hagiwara et al, 2004; Jenkins et al, 2009).

Figure EV1. Centriole in cilium formation.

1. Schematics of the centriole positioning during ciliogenesis in mammalian olfactory neurons.
2. Fluorescence time‐lapse images of the dynamic positioning of centrioles marked with SAS‐6::GFP (green) and of the plasma membrane and chromosome (red, labeled with mCherry) during cell division and neuritogenesis.
3. mMaple3::SAS‐4 localization in PQR neuron.
4. Schematic (left) and images (right) of the ciliated neurons in C. elegans. Red boxes represent the region of images shown in the left. Cell bodies of amphid neurons locate near the pharyngeal bulb, and cilia are positioned at the tip of dendrites that are extended anteriorly. Phasmid cilia are located at the tip of dendrites that were elongated toward the posterior of C. elegans. Yellow asterisks indicate the location of ciliary base.

Data information: Scale bars, 5 μm.

The nematode Caenorhabditis elegans offers a compelling model system to explore de novo ciliogenesis in the nervous system. All 60 ciliated cells in an adult C. elegans hermaphrodite are sensory neurons (Ward et al, 1975; Ware et al, 1975; Perkins et al, 1986; Inglis et al, 2007; Doroquez et al, 2014). In C. elegans embryos, previous live imaging studies have shown that sensory dendrites in amphids extend from the cell body via retrograde elongation, such that the growing dendrite is anchored at the presumptive nose while the soma translocates toward the posterior (Sulston et al, 1983; Heiman & Shaham, 2009; Schouteden et al, 2015). As a result, the basal bodies in these neurons do not undergo the dendritic translocation but are initially positioned at dendrite tips where sensory cilia emerge (Perkins et al, 1986; Heiman & Shaham, 2009). However, in mammalian OSNs, centrioles were detected in the middle of emerging dendrites and then arrived at the dendrite tips in mature neurons (Ying et al, 2014), which suggests that centriole translocation is an early step before the assembly of olfactory cilia (Cuschieri & Bannister, 1975a). Moreover, cilium formation in the C. elegans PQR sensory neuron may share similarities with ciliogenesis in mammalian OSNs (Figs 1A and EV1A). During the first larval stage, PQR is generated by asymmetric cell divisions of the QL neuroblast; after migration, PQR elongates its dendrite and constructs a sensory cilium at its dendrite ending (Sulston & Horvitz, 1977; Perkins et al, 1986). We have developed fluorescence time‐lapse microscopy to document Q neuroblast division, migration, and dendrite outgrowth in live C. elegans larvae (Chai et al, 2012).

Figure 1. Centriole migration and degeneration in the dendrite.

• A
  Schematics of the centriole positioning during ciliogenesis in Q neuroblast in Caenorhabditis elegans. Centriole translocates to the dendrite tip to initiate cilium assembly.
• B, C
  (B) Representative fluorescence time‐lapse images of the dynamic positioning of centrioles (green, labeled with SAS‐6::GFP) and the plasma membrane/chromosomes (red, labeled with mCherry) during ciliogenesis. Time 0 refers to when the leading edge starts to extend from the non‐motile cell body to form the dendrite, and the centriole migrates along with dendrite elongation. Note that centriole (1) translocates, whereas centriole (2) stays in the soma, and the centriole displacement defined by the position relative to cell body over time is quantified in (C) (Movie EV1). For details of the quantification, see methods and Fig EV2A; more examples are shown in Fig EV2B.
• D
  At the end of centriole migration, the centriole signal gradually decreases (left). Individual fluorescent centrioles were tracked, and intensity was quantified over time (right). More examples are shown in Fig EV2C.
• E
  No fluorescence of SAS‐4, SAS‐5, or SAS‐6 was detected at the base of the cilia in C. elegans sensory neurons in adult worms. CB, ciliary base, indicated with yellow asterisks.

Data information: Scale bars, 5 μm.

Here, we use the live imaging protocol, for the first time, to define the subcellular events involved in the formation of a sensory cilium of the C. elegans PQR neuron. We document centriole translocation from the cell body toward the dendrite tip when the dendrite elongates. Using biochemical and genetic approaches, we find that centriole translocation depends on cytoplasmic dynein‐1 and that the centriolar protein SAS‐5 interacts with the dynein light‐chain LC8. We further document the sequential accumulation of transition zone proteins and intraflagellar transport (IFT) proteins during the late stage of centriole translocation. Unexpectedly, centriolar proteins are removed upon the accumulation of transition zone proteins. Taken together, our results establish a comprehensive model of the de novo cilium formation in sensory neurons and provide insights into the same process in the mammalian olfactory system.

Results

Centriole translocation and degeneration in the dendrite

To monitor the dynamic behavior of centriole during ciliogenesis in C. elegans neurons, we labeled the components of the centriole core proteins including SAS‐4, SAS‐5, and SAS‐6 with a fluorescent protein (Leidel & Gonczy, 2003; Delattre et al, 2004; Leidel et al, 2005). Using a CRISPR‐Cas9‐assisted homologous recombination method, we constructed a mMaple3::SAS‐4 knock‐in (KI) strain. The fluorescence protein mMaple3 marks the tagged protein with green fluorescence (Wang et al, 2014). The KI animal is indistinguishable from the wild‐type (WT) nematodes in animal development. The particle bombardment transformation protocol (Praitis et al, 2001) was previously used to generate an integrant low‐copy GFP::SAS‐5 or SAS‐6::GFP strain. As illustrated in the early studies, these markers label centrioles during C. elegans development (Cabral et al, 2013); for example, the GFP fluorescence of SAS‐6::GFP was detected as a single spot at the spindle pole during Q cell asymmetric division and split to double spots (Fig EV1B). To visualize the PQR sensory neuron, we expressed mCherry‐tagged histone and plasma membrane proteins under the control of the Q neuroblast‐specific promoter Pegl‐17 (Figs 1B and EV1B; Ou et al, 2010).

To define the subcellular localization of the centriole during PQR development in live C. elegans larvae, we performed fluorescence time‐lapse microscopy of the developing larvae at 2‐min intervals over a period of 2–3 h. Our recordings usually start when the cell body of PQR arrives at the destination, while the leading edge continues to elongate to form the dendrite (Fig 1B). The dendrite grows out first, and the centriole translocation follows dendritic extension (Fig 1B). Time 0 refers to the start point of the leading edge extension from the non‐motile cell body. At approximately 15 min, one centriole moved into the growing dendrite, whereas the other remained immobile in the soma (Fig 1B and C, Movie EV1 and more examples in Fig EV2B). Quantification of the time‐lapse movies showed that the motile centriole translocated for 67 ± 22 min for a total distance of 5.1 ± 2.7 μm (Mean ± SD; N = 31). These results provide direct evidence in live animals that the centriole undergoes dendritic translocation to build the sensory cilium.

Figure EV2. Centriole translocation and degeneration.

• A
  Schematics of the quantification of centriole positioning.
• B, C
  Additional examples of centriole dynamic positioning (B) and decreased centriole signal (C).

Intriguingly, we found that the SAS‐6::GFP fluorescence of the motile centriole disappeared in the growing dendrite at the late stage of centriole translocation. A gradual decrease in SAS‐6::GFP fluorescence always occurred within 8 ± 2 min at 67 ± 22 min after the dendritic growth initiates, at the position of 5.1 ± 2.7 μm away from the cell body (Mean ± SD; N = 6; the representative images are shown in Figs 1D and EV2C), which suggests that the centriolar protein degradation may be programmed and argues against the possibility that they simply move out‐of‐focus during time‐lapse recording. We showed that the SAS‐6::GFP fluorescence from the non‐motile centriole in the cell body did not change during our recording and that the dendritic growth appeared to be normal (Figs 1B and EV1B), excluding the possibility that the loss of SAS‐6::GFP in the dendrite was caused by photo‐bleaching or photo‐toxicity. Other centriole markers including mMaple3::SAS‐4 could not be detected from the motile centriole in the dendrite but retained at the non‐motile centriole in the cell body (Fig EV1C). Consistently, all these centriolar proteins disappeared from the base of cilia of other sensory neurons in the C. elegans amphid and phasmid (Fig 1E). Thus, our live imaging data indicate that centriolar proteins are lost from the motile centriole in the growing dendrite and support the previous observation that the C. elegans centriole generally degenerates (Perkins et al, 1986; Dammermann et al, 2009; Williams et al, 2011).

Centriolar proteins are essential for ciliogenesis in Caenorhabditis elegans neurons

It is well established that the centriole‐derived basal body provides a template for nucleating axonemes. However, C. elegans centrioles do not have a cartwheel at their core but rather a central tube (O'Toole et al, 2003; Pelletier et al, 2006); and the early loss of the core centriolar proteins was detected in the developing dendrite (Fig 1D). We sought to rigorously examine the function of the centriolar protein SAS‐4, SAS‐5, and SAS‐6 during ciliogenesis in C. elegans sensory neurons. To bypass the crucial function of centriolar proteins in cell division and embryonic development, we used the somatic CRISPR‐Cas9 technique to generate conditional mutants of centriolar proteins in C. elegans L1 larvae (Fig 2A). We devised this technique to study the embryonically essential genes in Q cell division and migration (Shen et al, 2014). Using the same platform, we expressed the Cas9 endonuclease under the control of a heat‐shock‐inducible promoter to generate conditional mutants of sas‐4, sas‐5, and sas‐6 (Fig 2A and B). After heat‐shock induction of Cas9 expression, T7 endonuclease I (T7EI)‐based assays demonstrated that DNA fragments from sas‐4,‐5,‐6‐sg strains (‐sg for the conditional mutant) were digested into two small fragments of the expected sizes with the indels (insertion–deletion) of 9, 7, and 5%, respectively (Fig 2C). The conditional mutant embryos but not WT embryos exhibited embryonic lethality with the penetrance of 81 ± 4% (sas‐4‐sg), 94 ± 4% (sas‐5‐sg), and 85 ± 6% (sas‐6‐sg; Mean ± SD; N = 78–372; Fig 2D), which are consistent with the reported phenotypes of their RNAi animals (Gonczy et al, 2000; Sonnichsen et al, 2005). Thus, we constructed conditional mutations of embryonically essential centriole genes (sas‐4, sas‐5, and sas‐6).

Figure 2. Centriolar proteins are essential for ciliogenesis in Caenorhabditis elegans neurons.

• A
  Schematic representation of the cartwheel viewed from the proximal end. The nine symmetric cartwheel central hub is composed of SAS‐6 (blue), where SAS‐5 (red) localizes at the interior of centriole, and appears to function as a linker between SAS‐6 and SAS‐4 (green). SAS‐4 localizes more toward the periphery of the centriole, which associates with microtubules.
• B
  Gene models of centriolar proteins for generating conditional mutations; exons are shown as boxes, and arrows indicate sgRNA sequences corresponding to exons (Table EV1).
• C
  Representative gels showing the T7EI assay results for sas‐4‐sg, sas‐5‐sg, and sas‐6‐sg. PCR products amplified from the genomic DNA of worms expressing Phsp::Cas9 and PU6::sas‐4,5,6‐sg after heat‐shock treatment. Indels are indicated at the bottom.
• D
  Embryonic lethality was determined by quantifying viable embryos after heat‐shock treatment. N = 78–372; Mean ± S.D. (error bars). ****P < 0.0001 based on Student's t‐test. N.S., not significant.
• E
  Ciliary morphology (amphid, phasmid, and PQR neuron) in conditional mutants as visualized by OSM‐6::GFP. The penetrance was quantified at 0 ± 0% for WT (N = 62), 14 ± 7% for sas‐4‐sg (N = 76, P < 0.05), 23 ± 7% for sas‐5‐sg (N = 65, P < 0.01), and 18 ± 1% for sas‐6‐sg (N = 28, P < 0.001). Yellow asterisks indicate the location of ciliary base. Blue arrows mark the distribution of IFT particles along cilia.
• F
  Transition zone morphology in conditional mutants visualized by MKS‐5::mMaple3. The penetrance is 0 ± 0%, 12 ± 3%, 27 ± 10%, and 15 ± 1% for WT (N = 46), sas‐4‐sg (N = 69, P < 0.05), sas‐5‐sg (N = 98, P < 0.01), and sas‐6‐sg (N = 106, P < 0.001), respectively. Yellow asterisks indicate the location of ciliary base.
• G–I
  (G, H) Fluorescence time‐lapse images of MKS‐5::mMaple3 accumulation in sas‐5 mutant animals. Quantification is shown in (I), no accumulation (> 150 min), delayed accumulation (120–150 min), and normal (< 120 min) (N = 49) (Movie EV2).

Data information: Scale bars, 5 μm.

We next examined the ciliary morphology in these conditional mutant animals. Using an OSM‐6/IFT52::GFP reporter that marks an IFT‐particle subunit moving along the entire cilium, we found that the GFP fluorescence of OSM‐6::GFP was visible in the dendrite but was absent in 14 ± 7%, 23 ± 7%, and 18 ± 1% of sas‐4‐sg, sas‐5‐sg, and sas‐6‐sg animals, respectively (Mean ± SD; N = 28–76; the representative images are shown in Figs 2E and EV1D), indicating the lack of IFT in the absence of the centriolar proteins in amphid and phasmid cilia. The similar lack of OSM‐6::GFP fluorescence was also observed in cilia of the PQR neuron, with the penetrance of 15 ± 2%, 33 ± 11% and 30 ± 14% of sas‐4‐sg, sas‐5‐sg, and sas‐6‐sg (Mean ± SD; N = 12–20; the representative images are shown in Fig 2E). We further examined the transition zones in these mutant animals. The transition zone is a subdomain of the cilium characterized by Y‐shaped fibers connecting the doublet microtubules and ciliary membrane and functioning as a ciliary gate to regulate IFT (Williams et al, 2011). We generated a mMaple3 knock‐in animal of MKS‐5 (Meckel‐Gruber syndrome) protein to visualize the transition zone and mMaple3‐tagged MKS‐5 localized at the ciliary base in 100% of the WT adult animals. In consistent with the absence of OSM‐6::GFP in cilia, no fluorescence of MKS‐5::mMaple3 could be detected at the base of cilia, with the penetrance of 12 ± 3%, 27 ± 10%, and 15 ± 1% in amphid and phasmid cilia of the adult sas‐4‐sg, sas‐5‐sg, or sas‐6‐sg animals, respectively (Mean ± SD; N = 46–106; the representative images are shown in Fig 2F). The TZ defects were also detected in PQR neuron as 18 ± 2%, 40 ± 1%, and 17 ± 4% of sas‐4‐sg, sas‐5‐sg, or sas‐6‐sg animals were lack of MKS‐5::mMaple3 fluorescence intensity around their transition zones (Mean ± SD; N = 17–67; the representative images are shown in Fig 2F). We next examined the MKS‐5 accumulation in L1 larvae in the absence of the core centriole proteins. Our live imaging analysis showed that the MKS‐5::mMaple3 fluorescence was enriched as puncta in the dendrite 68 ± 23 min after the dendrite elongates in WT animals (Fig 5A; N = 54). By contrast, MKS‐5::mMaple3 did not form any puncta in 29% of sas‐5‐sg conditional mutants after the time‐lapse recording for 2–3 h (Mean ± S.D.; N = 49; the representative images and movie are shown in Fig 2G and I, and in Movie EV2). Although 24% of the sas‐5‐sg animals eventually formed the MKS‐5 fluorescence puncta at the dendrite tip, the process was severely retarded, taking 125 ± 33 min (Mean ± SD; N = 49; the representative images and movies are shown in Fig 2H and I, and in Movie EV2). The delayed formation of MKS‐5 puncta in 24% of sas‐5‐sg conditional mutants is likely because residual SAS‐5 builds a deformed centriole that is not as good at templating the cilium. Conditional mutations of sas‐5 may not completely remove SAS‐5 as it is essential for cell division and other events before ciliogenesis, and the absence of GFP::SAS‐5 fluorescence in sas‐5‐sg animals may not indicate the complete loss of SAS‐5, making it difficult to examine the role of residual SAS‐5 in partial ciliogenesis. We cannot exclude the possibility that cells may form cilia without centrioles but not very efficiently. Together, these data indicate that centriolar proteins are essential for ciliogenesis in C. elegans.

Figure 5. The sequential recruitment of transition zone and IFT proteins.

1. MKS‐5 accumulation during ciliogenesis. Arrow indicates the MKS‐5 punctum (Movie EV2).
2. The OSM‐6 accumulation. Arrow indicates OSM‐6 punctum (Movie EV2).
3. Quantification and comparison of the time for SAS‐6 loss, MKS‐5 accumulation, and OSM‐6 accumulation. SAS‐6 disappears at 65 ± 22 min after time 0 (N = 53). The MKS‐5 puncta appear at 68 ± 23 min (N = 54, N.S. vs. SAS‐6 loss), and OSM‐6 accumulates at 102 ± 29 min (N = 27, ****P < 0.0001 vs. SAS‐6 loss). N.S., not significant.
4. A representative fluorescence accumulation of MKS‐5 (left) and a trace of the fluorescence increase fitted with a second‐order polynomial (quadratic) model (right). More traces are in Fig EV5A.
5. A representative fluorescence accumulation of OSM‐6 (left) and a trace of the fluorescence increase fitted with an exponential growth model (right). More traces are in Fig EV5B.
6. Time‐lapse images of centriole degeneration (SAS‐6::GFP), IFT accumulation (OSM‐6::mCherry), and the PQR cytosol (Plin‐32::TagBFP) (Movie EV5).
7. Quantification of the time sequences: 76 ± 23 min for SAS‐6 loss and 128 ± 17 min for OSM‐6 accumulation (N = 8, ***P < 0.001).
8. Time‐lapse images of transition zone formation (MKS‐5::mMaple3), IFT accumulation (IFT‐81::BFP), and the PQR membrane (Pegl‐17::myri::TagRFP) (Movie EV5).
9. Quantification of the time sequences: 63 ± 24 min for MKS‐5 accumulation and 95 ± 28 min for IFT‐81 accumulation (N = 8, ***P < 0.001).

Data information: Mean ± SD; Student's t‐test; Scale bars, 5 μm.

Centriole translocation depends on the cytoplasmic dynein‐1

We next investigated the mechanism underlying centriole translocation along the dendrite. To explore microtubule polarity in the developing dendrites of L1 larvae, we constructed the microtubule plus‐end binding protein EBP‐2::GFP knock‐in strain. The fluorescence of EBP‐2::GFP appeared as comet‐like streaks in the C. elegans embryos and in adult somatic cells (Movie EV3). As shown in the kymograph in Fig 3A, 80% of the EBP‐2::GFP comets moved away from the growth cone to the cell body at 0.22 ± 0.11 μm/s, whereas the other 20% of the EBP‐2::GFP comets moved in the opposite direction with a speed of 0.22 ± 0.11 μm/s, indicating that dendritic microtubules mostly emanate from the tip of the dendrite (Mean ± SD; N = 223 tracks in 12 cells; the representative images are shown in Figs 3A and EV3A). Consistently, in sensory neurons of adult C. elegans, 94% of EBP‐2::GFP moved from the cilium base toward the cell body (Hao et al, 2011). The orientation of microtubules and the unidirectional movement of centrioles along the dendrite suggest that a minus‐end‐directed microtubule motor protein might be involved. Importantly, by visualizing DHC‐1 localization in a GFP::DHC‐1 animal, we showed that DHC‐1 was restricted within the dendrite and did not enter cilia (Fig 3B). We postulate that centriole translocation may be driven by dynein‐1 along the dendritic microtubules.

Figure 3. Centriole translocation requires cytoplasmic dynein‐1.

• A
  EBP‐2::GFP protein dynamics in PQR neuron during ciliogenesis in L1 larva (top). A kymograph (left bottom) with corresponding cartoons (right bottom) showing 80% of EBP‐2::GFP comets move from cilium to soma at the speed of 0.22 ± 0.11 μm/s (N = 223) (Movie EV3).
• B
  Dynein motility in dendrites of ciliated neurons. A kymograph (right top) with corresponding cartoons (right bottom) showing the soma‐to‐cilium directed movement of dynein; scale bars, 2 μm, 10 s.
• C
  Centriole movement defects in conditional mutants visualized by SAS‐6::GFP; 21 ± 11% for dhc‐1‐sg (N = 152, P < 0.01), 11 ± 5% for dlc‐1‐sg (N = 92, P < 0.01), and 10 ± 5% for lis‐1‐sg (N = 76, P < 0.05).
• D–F
  (D, F) Fluorescence time‐lapse images and quantification of centriole positioning during ciliogenesis in Dynein‐1 mutants. Among these cells, 30% show two immobile centrioles remaining in the soma (left), 9% have one centriole in the dendrite without continuous movement toward the dendrite tip (middle), and 4% contain centriole that underwent backward migration to the cell body (right) (N = 46). Note that at 0 min dendrite already extended more than 1 μm (Movie EV4). (E) The centriole displacement tracks in Dynein‐1 mutants compared with wild‐type animals; the quantification is shown in (F).

Data information: Scale bars, 5 μm except for Fig 3B.

Figure EV3. Dynein‐1 is involved in regulating centriole translocation.

1. EBP‐2::GFP protein dynamics in ciliated neurons in adult worms. A kymograph (middle) with corresponding cartoons (bottom) showing 100% of EBP‐2::GFP comets moving from the cilium to the soma.
2. Fluorescence time‐lapse images and quantification of centriole positioning during ciliogenesis in dynein light‐chain 8 mutants.
3. Fluorescence time‐lapse images and quantification of centriole positioning during ciliogenesis in lis‐1‐sg mutants.
4. A gel showing the T7EI assay results for dhc‐1‐sg.
5. Ciliary morphology (amphid, phasmid, and PQR neuron) in Dynein subunits conditional mutants visualized by OSM‐6::GFP. The cilium length was reduced from 8.3 ± 0.2 μm in WT animals to 3.7 ± 0.4 μm in dhc‐1‐sg, 3.5 ± 0.4 μm in dlc‐1‐sg, or 0.6 ± 0.4 μm in lis‐1‐sg conditional mutants (N = 25–61, Mean ± SEM). Yellow asterisks indicate the location of ciliary base.

Data information: Scale bars, 5 μm.

To study the function of dynein‐1 in centriole translocation, we examined the position of centrioles of PQR neuron in the dhc‐1‐sg conditional mutant strain. At the late L1 larval stage, centriole translocation was completed in the dendrite of WT PQR neurons. However, in 21 ± 11% of dhc‐1‐sg mutant animals (N = 152) that properly elongated their dendrites, no centrioles could be detected in the dendrite; instead, two centrioles were retained in the cell body. Conditional mutations of the dynein‐1 light‐chain LC8 or the associated subunit Lissencephaly protein LIS‐1 caused the same defects of centriole translocation, with penetrance of 11 ± 5% for dlc‐1‐sg (N = 92) and 10 ± 5% for lis‐1‐sg (N = 76; Fig 3C). To address which aspects of centriole movement require dynein‐1, we followed centriole behavior in dynein‐1 conditional mutant animals. For these live cell imaging analyses, our time‐lapse recording started when the dendrite was over 1.5 μm in length. Three patterns of centriole behavior indicated the effects of dynein‐1 inhibition. In one pattern, 30% of cells retained two immobile centrioles in the soma, while the dendrite extended; in the second pattern, 9% of cells had one centriole in the elongating dendrite but this centriole did not show any continuous movement toward the dendrite tip; in the third pattern, centriole neared the dendrite tip in 4% of cells but underwent backward movement to the soma (Fig 3D–F; N = 46). These phenotypes were reproducibly observed in dlc‐1‐sg (22% for staying in soma; 7% for backward movement; N = 27) and in lis‐1‐sg conditional mutant animals (9% for non‐motile in the dendrite; 9% for backward movement; N = 11; the representative images and movie are shown in Figs 3E and F, and EV3B and C; Movie EV4). The conditional mutations of dlc‐1 or lis‐1 reduced ciliary length and disrupted the ability of these cilia to uptake fluorescence dye (i.e., the Dyf phenotype; Li et al, 2015). Consistently, we found that the lack of DHC‐1 caused Dyf phenotype in 23 ± 5% animals (N = 165 from three generations, Mean ± SEM). Importantly, the cilium length was significantly reduced from ~8 μm in WT animals to 3.7 ± 0.4 μm in dhc‐1 conditional mutant worms (N = 48, Mean ± SEM; the representative images are shown in Fig EV3D and E). Overall, our results indicate that the cytoplasmic dynein‐1 is required for centriole translocation along the growing dendrite.

LC8 and SAS‐5 bind in vitro and colocalize in vivo

We next sought to understand how dynein‐1 is involved in the translocation of centriolar proteins. The recent structural study provided evidence that dynein light‐chain LC8 dimers bind to the Drosophila Ana2/SAS‐5 (Slevin et al, 2014). Using an anti‐GFP antibody, we purified GFP‐tagged SAS‐5 with its associated proteins from the lysate of transgenic C. elegans larvae that expressed GFP::SAS‐5 (Fig 4B). Protein constituents were determined by means of liquid chromatography–tandem mass spectrometry. Two interesting hits from the purification were the dynein light‐chain DLC‐1/LC8 and heavy‐chain DHC‐1 from the dynein‐1 complex (Fig 4A and B; Datasets EV1 and EV2). To examine the interactions between DLC‐1 and SAS‐5, we performed a GST pulldown assay using the purified GST‐DLC‐1 and the worm lysate containing GFP::SAS‐5. As shown in Fig 4C, GFP::SAS‐5 was detected to associate with GST‐DLC‐1 but not GST control. We further demonstrated a direct interaction between DLC‐1 and SAS‐5 using the pulldown assay with the purified GST‐DLC‐1 and His‐GFP‐SAS‐5 (Fig 4D). Moreover, our GFP‐affinity purification of binding partners of SAS‐6 also identified components of the dynein‐1 motor complex (Fig EV4A and B; Datasets EV1 and EV2).

Figure 4. Dynein light‐chain LC8/DLC‐1 interacts with SAS‐5 in vitro and colocalizes in vivo .

1. Mass spectrometry analysis of the indicated affinity purifications. The percentage of sequence coverage was listed.
2. Western blot with anti‐GFP antibody showing the immunoprecipitation of GFP::SAS‐5 from C. elegans using GFP‐TrapA beads (left). The mass spectrum of a unique peptide NADMSDDMQQDAIDCATQALEK in DLC‐1 (LC8).
3. SAS‐5 interacts with DLC‐1. GST pulldown assay using GST‐DLC‐1 and worm lysate from the strain expressing GFP::SAS‐5.
4. Direct interaction between SAS‐5 and DLC‐1 by GST pulldown assay using GST‐DLC‐1 and His‐GFP‐SAS‐5.
5. Schematic representation showing asymmetric cell division to generate PQR and neuritogenesis of PQR to form the cilium (left). Fluorescence time‐lapse images of GFP‐tagged DLC‐1 localization during cell division and neuritogenesis (right) (Movie EV3). Scale bar, 5 μm.

Figure EV4. Dynein light‐chain LC8/DLC‐1 interacts with the SAS‐6 in vitro .

1. Mass spectrometric analysis of the indicated affinity purifications, listing the percentage sequence coverage.
2. Western blot with anti‐GFP antibody shows the immunoprecipitation of SAS‐6::GFP from worms using GFP‐TrapA beads (left). The mass spectrum of the unique peptide YNIEKDIAAYIK in DLC‐1 (LC8).
3. Fluorescence time‐lapse images of GFP‐tagged DLC‐1 localization during ciliogenesis. Scale bar, 5 μm.

Given the above genetic and biochemical observations, we examined where DLC‐1 might regulate centriole translocation. To this end, we constructed a DLC‐1::GFP knock‐in strain to visualize the localization of the dynein light‐chain LC8 during Q neuroblast development. Live imaging analysis showed that the DLC‐1::GFP fluorescence was enriched as puncta at the spindle pole where centrioles localize during Q cell division. After Q cell division and migration, one of the two GFP puncta entered the growing dendrite and underwent the directional movement toward the dendritic tip as the motile centriole at the velocity of 0.11 ± 0.03 μm/min (Figs 4E and EV4C; Movie EV3). Our data suggest that the dynein‐1 powers centriole translocation likely via the direct interaction between LC8 and SAS‐5.

The sequential recruitment of transition zone and IFT proteins

We next examined the sequence of the recruitment of transition zone proteins and IFT components in relation to the loss of centriolar proteins. Using an MKS‐5::mMaple3 knock‐in strain, we showed that the green fluorescence of MKS‐5::mMaple3 signal was first detected at 68 ± 23 min after the dendrite started to grow (Fig 5A and C; Movie EV2; Mean ± SD, N = 54). A second‐order polynomial (quadratic) model appeared to fit the increase in fluorescence over time from the MKS‐5::mMaple3 spot (Figs 5D and EV5A). For the IFT components, the OSM‐6::GFP fluorescence emerged approximately 102 ± 29 min after the initiation of dendritic growth (Fig 5B and C; Movie EV2; Mean ± SD, N = 27), and an exponential growth equation model was applied to fit the increase in the fluorescence of OSM‐6::GFP over time (Figs 5E and EV5B). These observations reveal the different timings and patterns of protein enrichment in the formation of the transition zone and cilium. Our comparison of the time points of centriolar protein elimination and transition zone protein accumulation did not reveal any significant differences (P = 0.85, t‐test), suggesting that centriolar proteins were removed when the transition zone started to form. By contrast, IFT‐protein accumulation was significantly later than either centriolar protein elimination or the transition zone formation (P < 0.0001; Figs 5C and EV5C).

Figure EV5. The centriole is required for transition zone formation and IFT accumulation.

1. Fluorescence accumulation traces of MKS‐5 fitted with a second‐order polynomial (quadratic) model.
2. Fluorescence accumulation traces of OSM‐6 fitted with an exponential growth equation model.
3. Comparison of the times for SAS‐6 loss, MKS‐5 accumulation, IFT81 accumulation, and OSM‐6 accumulation. **P < 0.01, ****P < 0.0001 based on Student's t‐test; N.S., not significant.
4. Hydrolethalus syndrome protein 1 (HYLS‐1) persists throughout mitosis and ciliogenesis. At time 0, dendrite elongation is completed. Scale bars, 5 μm.

We conducted triple‐fluorescence live imaging analysis to further confirm the above sequences. For these experiments, we generated the fluorescent animals expressing SAS‐6::GFP OSM‐6::mCherry, and TagBFP‐labeled cytosol of PQR. The fluorescence of SAS‐6::GFP disappeared at 76 ± 23 min after the dendritic growth, and OSM‐6::mCherry fluorescence was not enriched at the dendrite tip until 49 ± 12 min after the loss of SAS‐6::GFP (Fig 5F and G; Movie EV5; Mean ± SD; N = 8, P < 0.001). In an animal expressing MKS‐5::mMaple3 KI, IFT‐81::TagBFP KI and mCherry‐labeled PQR plasma membrane and histone, MKS‐5 in the transition zone (63 ± 24 min) was enriched 32 ± 11 min earlier than the accumulation of IFT‐81 protein (95 ± 28 min; Fig 5H and I; Movie EV5; Mean ± SD; N = 8). Together, our data show that centriolar proteins are lost at around the same time as the formation of the transition zone but earlier than the enrichment of IFT components at the dendrite tip.

Discussion

In summary, this study has defined the subcellular events during neuronal ciliogenesis in a developing sensory neuron in C. elegans. We provide evidence that dynein‐dependent centriole translocation, elimination of centriole subunits, and the sequential recruitments of transition zone and IFT components are involved in this process. We suggest that the formation of olfactory cilia in mammalian OSNs may use mechanisms similar to those delineated here.

How a centriole imparts the ninefold symmetry information to template axonemal microtubules during ciliogenesis has been a long‐standing question in developmental neurobiology. Axonemal microtubules are invariably organized in a radial ninefold symmetry, which is thought to originate from the nine spokes in the cartwheel. Centriole satellites are essential for ciliogenesis (Anderson, 1972; Vorobjev & Chentsov Yu, 1982; Tanos et al, 2013; Tateishi et al, 2013); the C. elegans homolog of hydrolethalus syndrome protein 1 (HYLS‐1) is required for anchoring of centrioles at the plasma membrane in sensory neurons (Dammermann et al, 2009; Wei et al, 2016). HYLS‐1 marks the centriolar MTs that spread out at the base of worm cilia, presumably due to the loss of the central tube (Dammermann et al, 2009; Doroquez et al, 2014). Consistent with this notion, we show that HYLS‐1 localizes at the centriole and persists at the base of the mature cilia (Fig EV5D). The central tube may become unnecessary for the subsequent formation of the transition zone or recruitment of ciliary components, which may explain the elimination of the centriole tube before the recruitment of IFT subunits.

Why does the centriole degenerate during ciliogenesis? Previous TEM results have shown that centriole degeneration appears to occur shortly after the onset of ciliogenesis during late embryogenesis (White et al, 1986; Nechipurenko et al, 2017; Serwas & Su, 2017), which is consistent with the loss of GFP fluorescence of SAS‐6 around transition zone formation (Fig 1B). During the mammalian cell cycle, SAS‐6 protein is degraded at the end of mitosis and accumulates again beginning in the late G1 phase, and the ectopic higher levels of SAS‐6 promote the formation of excess procentrioles or centrioles (Strnad et al, 2007; Arquint et al, 2012). The elimination of centriolar proteins may prevent the formation of excess centrioles, ensuring the correct cilium number in a given cell. Although the basal bodies retain at the base of the cilium in mammalian OSNs (Falk et al, 2015), it remains to be examined whether cartwheel proteins such as SAS‐6 is localized in these basal bodies, leaving it an open question whether the mammalian centriolar proteins are removed, like their C. elegans homologs, during olfactory ciliogenesis.

How are centriolar proteins lost in the developing dendrite? In a mitotic cell, APC^Cdh1 targets SAS‐6 to the 26S proteasome for degradation (Arquint et al, 2012). Recent studies have uncovered the functional interactions between autophagy and ciliogenesis: Removal of OFD1 by autophagy at centriolar satellites promotes ciliogenesis, whereas the degradation of IFT proteins by autophagy regulates ciliary length (Tang et al, 2013). Centriole degeneration during ciliogenesis may also require the contribution of the 26S proteasome or autophagy. However, the SAS‐6::GFP fluorescence persists throughout the cell cycle in the Q neuroblast lineages (Figs 1B and EV1B), suggesting that centriole degeneration in C. elegans may utilize additional mechanisms. The motile centriole in the dendrite degenerates, whereas the non‐motile centriole in the cell body remains, suggesting that the degradation activity is tightly controlled based on the subcellular location or that two centrioles show distinct responses to degeneration signal. In support of the distinct fates of these two centrioles, the immotile centriole also has centriolar proteins that interact with dynein‐1 and DLC‐1; however, its motility could be barely detected during PQR development (Figs 1C and D, and 4E).

This work uncovers an important role of cytoplasmic dynein‐1 in centriole translocation along the dendrite. We identify a direct interaction between the dynein light‐chain LC8 and SAS‐5 and suggest that this interaction may couple dynein‐1 with the centriole. This interaction pair was previously reported from the Drosophila neuroblast asymmetric division (Wang et al, 2011). Crystallographic studies showed that four LC8 dimers bind to the central region of the Drosophila Ana2/SAS‐5 and that these LC8‐binding sites are not readily apparent in the C. elegans SAS‐5 (Slevin et al, 2014). Our biochemical data revealed that the interaction between DLC‐1 and SAS‐5 is conserved in C. elegans. It remains unclear how LC8 bridges the dynein motor complex to the centriolar protein complex; further, some species possess LC8 but do not have the dynein heavy chain, suggesting that LC8 can localize with the centriole hub in a complex with SAS‐5, independent of the dynein motor (King, 2008; Lightcap et al, 2008). However, our mass spectrometric analysis of the binding partners of GFP::SAS‐5 detected both LC8 and the dynein motor; importantly, conditional mutations of the dynein motor and LC8 caused the same defects on centriole translocation, indicating that the dynein‐1 motor and LC8 function together to transport the centriole. Cytoplasmic dynein‐1 is the only microtubule minus‐end motor that we examined in this study, and we cannot exclude the possibility that other minus‐end kinesin motors are involved in centriole translocation. The speed of centriole translocation along the dendrite is 0.08 ± 0.01 μm/min (N = 31); intriguingly, the centriole was moved by dynein‐1 at a relatively constant rate of 0.1 μm/min in migrating neural precursors of the mouse brain tissue (Tsai et al, 2007). Despite the similar translocation speeds of the C. elegans centriole and the mammalian centrosome, the cytoplasmic dynein‐1 can move at > 1 μm/s in live cells (Pilling et al, 2006; Reck‐Peterson et al, 2006; Moughamian & Holzbaur, 2012; Maday et al, 2014). As shown in the kymograph of GFP::DHC‐1 along the dendrite, dynein‐1 moves at 0.18 ± 0.09 μm/s in C. elegans neurons (Mean ± SD; N = 73; the representative images are shown in Fig 3B). The speed discrepancy may result from the transient interaction between centriole and dynein or multiple mechanisms may be involved at different stage of centriole translocation.

Transmission electron microscopy studies revealed two general phases of ciliogenesis. The early phase includes the binding of ciliary vesicles to the basal body, after which the transition zone emerges and matures through fusion of secondary vesicles until the ciliary gate is formed. The late phase includes axoneme extension in which ciliary precursors are delivered into the growing cilia by IFT‐protein machinery (Sorokin, 1962; Reiter et al, 2012). In addition to the shared steps, different sensory neurons can apply distinct mechanisms to position their centrioles. Basal bodies were initially located at the dendrite tips in sensory neurons of C. elegans embryos (Dammermann et al, 2009; Schouteden et al, 2015), whereas the centriole in PQR translocates to the dendrite tip. Future studies may uncover other fascinating mechanisms of ciliogenesis from the mammalian olfactory system that consists of millions of OSNs.

Materials and Methods

Caenorhabditis elegans strains, genetics and DNA manipulations

Caenorhabditis elegans strains were cultivated on NGM plate seeded with the Escherichia coli strain OP50 at 20°C according to standard protocols. To generate conditional knockout mutants, CRISPR‐Cas9 plasmid construction and injection was carried out as described previously (Li et al, 2015). For transgenic expression, DNA plasmids were microinjected to the germline at 10–50 ng/μl with co‐injection markers pRF4 [rol‐6 (su1006)] and/or Podr‐1::dsRed into N2 or co‐injection the unc‐76‐rescuing plasmid p76‐16B into unc‐76 (e911) animals. To express mMaple3 in C. elegans, we reengineered it by changing its codon use and inserting artificial introns according to the original sequences (Green et al, 2008; Wang et al, 2014). For knock‐in repair template, briefly, GFP‐ or mMaple3‐fluorescent protein gene flanked by 1‐1.5kb upstream and downstream homologous arms of target gene was inserted into pPD95.77 via In‐Fusion Advantage PCR Cloning Kit (Clontech). Knock‐in strains were selected and confirmed by PCR. A complete list of strains, primers, and plasmids can be found in the Tables EV1 and EV2.

Heat‐shock treatment and viability

Approximately 100 eggs that were laid after 4–6 h were collected and treated with heat‐shock (33°C for 1 h) to generate conditional mutation. For viability assay, eggs survived from heat‐shock treatment were grown into young adults; viability is defined as the hatched worms divided by the number of the original eggs. All data were collected from more than three generations.

Molecular assay

The detection of molecular lesions induced by conditional knockout via CRISPR‐Cas9 strategy followed previous methods (Shen et al, 2014). Genomic DNA was isolated from heat‐shock‐treated worm lysis, the target fragments were amplified by PCR, and purified DNA was digested by 5 units of T7 endonuclease I (New England Biolabs) at 37°C for 30 min. Indel percentage was determined by the formula 100× (1 − (1 − (b + c)/(a + b + c))^1/2), where a is the integrated intensity of the intact band and b and c are the integrated intensities of digested small bands.

Live cell imaging

Caenorhabditis elegans L1 larvae were anesthetized with 1 mg/ml levamisole in M9 buffer, mounted on 3% agarose pads, and maintained at 22°C (Chai et al, 2012). Imaging system includes an Axio Observer Z1 microscope (Carl Zeiss MicroImaging, Inc.) equipped with a 100X, 1.45 N.A. objective, an EM CCD camera (Andor iXon+ DU‐897D‐C00‐#BV‐500), and the 488‐nm and 568‐nm lines of a Sapphire CW CDRH USB Laser System to a spinning disk confocal scan head (Yokogawa CSU‐X1 Spinning Disk Unit). Time‐lapse images were acquired with exposure time of 200 ms with 3 μm Z stack projection at every 2 min for imaging the entire ciliogenesis process. Images were acquired by μManager (www.micro-manager.org), and images were processed and quantified by ImageJ software (http://rsbweb.nih.gov/ij/).

Quantifications and statistical analysis

To quantify the centriole translocation during PQR ciliogenesis, the PQR is rotated to the anterior–posterior direction, the time zero of PQR ciliogenesis was defined as the completion of cell migration of the PQR cell. The centriole position is digitalized as the XY coordinate, and the parameters such as distance and time were calculated by the XY coordinate of each time point relative to the soma (Fig EV2A). The formation of transition zone and IFT proteins were determined by the appearance of the fluorescence spot at the dendrite tip, and the time gap between two processes was compared and calculated. Student's t‐test and chi‐square analysis were used to examine statistical differences of ciliary phenotypes between WT and mutants as indicated in the figure legends, and two‐sided P‐value was shown.

Protein purification

We amplified sas‐5 and dlc‐1 cDNA from a C. elegans cDNA library. Full‐length SAS‐5 tagged with GFP and DLC‐1 were cloned in pET15a and pGEX4T vectors, respectively. The resultant plasmids were transformed to E. coli strain BL21 (DE3) and grown at 37°C in LB media to produce His‐GFP‐SAS‐5 and GST‐DLC‐1 protein. Protein expression was induced at OD600 0.6 by isopropyl‐β‐D‐1‐thiogalactopyranoside (IPTG, 0.3 mM final concentration) at 16°C. After ∼16 h, E. coli were harvested and disrupted by sonication in the lysis buffer [25 mM Tris‐HCl (pH 7.5), 150 mM NaCl, 0.5% Nonidet P40, 10% glycerol] supplemented with protease inhibitor cocktail (cOmplete, Roche, Basel, Switzerland). After centrifugation at 40,000 g for 30 min, cell lysate of His‐GFP‐SAS‐5 was incubated with nickel beads for 60 min at 4°C. Resin beads were washed with lysis buffer supplemented with 20 mM imidazole, and protein was eluted with lysis buffer supplemented with 500 mM imidazole followed by gel filtration using the chromatography system (BioLogic DuoFlow; Bio‐Rad Laboratories) with a Superdex 200 10/300 GL column (GE Healthcare; equilibrated with lysis buffer containing 1 mM DTT). Cell lysate of GST‐DLC‐1 was incubated with glutathione Sepharose beads for 60 min at 4°C. Beads were washed with PBS containing 250 mM NaCl and re‐equilibrated in PBS supplemented with 20% glycerol and were flash‐frozen in liquid nitrogen for pulldown assay.

Pulldown assay

2 ml of worm pellet expressing GFP::SAS‐5 was mixed with 2 ml lysis buffer containing protease inhibitors and 4 ml of 0.5‐mm‐diameter glass beads, and was disrupted with FastPrep‐24 (MP Biomedicals) at 6.5 m/s, 15 s/pulse × 5 pulses with 5‐min intervals on ice. Lysates were clarified by centrifugation at 12,000 g for 30 min and were mixed with the beads associated with 30 μg GST‐DLC‐1 or GST for 60 min at 4°C. The beads were washed with lysis buffer supplemented with 250 mM NaCl and resuspended with a SDS sample buffer followed by SDS‐PAGE. For pulldown assays using the purified proteins, 10 μg His‐GFP‐SAS‐5 was incubated with beads associated with 30 μg GST‐DLC‐1 or GST protein. The reaction was performed in 50 μl lysis buffer for 60 min at 4°C, and the beads were washed and resuspended in a SDS sample buffer followed by Western blot analysis with anti‐GFP antibody (Roche,# 11814460001,RRID:AB_390913).

Mass spectrometry analysis

GFP::SAS‐5 and SAS‐6::GFP mass spectrometry was performed as previously described (Falkner & Andrews, 2005). Worms were lysed and clarified with same protocol in pulldown assay. Cleared worm extracts were incubated with the GFP‐Trap A beads (Chromoteck) for 60 min and washed three times with lysis buffer. Immunoprecipitates were dissolved from the GFP‐IP beads in 100 mM Tris, pH 8.5 and 8 mol/l urea. Proteins were reduced with 5 mM TCEP and alkylated with 10 mM Iodoacetamide followed by digestion with trypsin at 37°C for 14 h. The resulting peptides were analyzed twice on a Q Exactive mass spectrometer (ThermoFisher Scientific) interfaced with an Easy‐nLC1000 liquid chromatography system (ThermoFisher Scientific). Peptides were loaded on a 75 μm × 4 cm trap column packed with 10 μm, 120 Å ODS‐AQ C18 resin (YMC Co., Ltd. Kyoto, Japan) and connected to a 75 μm × 10 cm analytical column packed with 1.8 μm, 100 Å Luna C18 resin (Welch Materials, Shanghai, China). After desalting, peptides were separated at a flow rate of 200 nl/min with a 79‐min linear gradient from 5% buffer B (100% ACN, 0.1% FA), 95% buffer A (0.1% FA) to 28% buffer B followed by a 10‐min gradient to 80% buffer B, 3‐min gradient to 100% buffer B and maintaining at 100% buffer B for 10 min. The top 15 most abundant precursor ions from the survey scan were selected for higher‐energy collisional dissociation (HCD); R = 70,000 in full scan, R = 17,500 in HCD scan; AGC targets were 3e6 for FTMS full scan, 1e5 for MS2; minimal signal threshold for MS2 = 5e4; +1 and unassigned precursors are excluded; normalized collision energy is 27 for HCD; dynamic exclusion is 30 s; ion 445.12003 is used for internal calibration. Raw data were extracted using the RawXtract software. The MS/MS spectra were searched against the C. elegans protein database using with Prolucid (Xu et al, 2015) and the search results of the two technical replicates were filtered separately using DTASelect 2 (Tabb et al, 2002) by requiring 1% FDR at the peptide level, precursor mass accuracy ≤5 ppm, Z score ≥ 4, and a minimum of one peptide for a protein ID. The FDR rates at the protein level were 0.61 and 1.76%, respectively. The complete MS dataset is provided in Datasets EV1 and EV2.

Author contributions

GO and WJL conceived the experiments; WJL, PY, ZZ, XZ, and WL performed the experiments; and GO and WJL wrote the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Table EV1

Table EV2

Dataset EV1

Dataset EV2

Movie EV1

Movie EV2

Movie EV3

Movie EV4

Movie EV5

Review Process File

The EMBO Journal (2017) 36: 2553–2566