Tissue-specific degradation of essential centrosome components reveals distinct microtubule populations at microtubule organizing centers

Non-centrosomal microtubule organizing centers (ncMTOCs) are found in most differentiated cells, but how these structures regulate microtubule organization and dynamics is largely unknown. We optimized a tissue-specific degradation system to test the role of the essential centrosomal microtubule nucleators γ-tubulin ring complex (γ-TuRC) and AIR-1/Aurora A at the apical ncMTOC, where they both localize in Caenorhabditis elegans embryonic intestinal epithelial cells. As at the centrosome, the core γ-TuRC component GIP-1/GCP3 is required to recruit other γ-TuRC components to the apical ncMTOC, including MZT-1/MZT1, characterized here for the first time in animal development. In contrast, AIR-1 and MZT-1 were specifically required to recruit γ-TuRC to the centrosome, but not to centrioles or to the apical ncMTOC. Surprisingly, microtubules remain robustly organized at the apical ncMTOC upon γ-TuRC and AIR-1 co-depletion, and upon depletion of other known microtubule regulators, including TPXL-1/TPX2, ZYG-9/ch-TOG, PTRN-1/CAMSAP, and NOCA-1/Ninein. However, loss of GIP-1 removed a subset of dynamic EBP-2/EB1–marked microtubules, and the remaining dynamic microtubules grew faster. Together, these results suggest that different microtubule organizing centers (MTOCs) use discrete proteins for their function, and that the apical ncMTOC is composed of distinct populations of γ-TuRC-dependent and -independent microtubules that compete for a limited pool of resources.


Introduction
Described nearly 50 years ago, microtubule organizing centers (MTOCs) generate specific spatial patterns of microtubules as needed for cell function [1].
The best-studied MTOC is the centrosome, a non-membrane bound organelle that organizes microtubules into a radial array from its pericentriolar material (PCM) or from subdistal appendages attached to the mother centriole. However, in many types of differentiated cells, microtubules are organized at noncentrosomal sites to accommodate diverse cell functions. In animal cells, these non-centrosomal MTOCs (ncMTOCs) can be found in the axons and dendrites of neurons, around the nuclear envelope of skeletal muscle cells, at the apical surface of epithelial cells, and at the Golgi complex [2][3][4][5][6][7][8]. ncMTOCs can promote non-radial arrangements of microtubules, such as the linear arrays of microtubules present along the apico-basal axis in epithelial cells. How ncMTOCs are established and whether they are composed of the same proteins that impart MTOC activity at the centrosome is largely unknown.
In general, MTOCs can be defined as cellular sites that nucleate, anchor, and stabilize microtubules; however, the molecular basis for these functions has been elusive [9,10]. Because of the inherent structural and chemical polarity of microtubule polymers, microtubules are nucleated and anchored at their minus ends. Thus, a defining feature of MTOCs is that they interact with microtubule minus ends. The first microtubule minus-end protein described was γ-tubulin, which together with GCP2 and GCP3, forms the γ-tubulin small complex (γ-TuSC) [11]. In some organisms, additional γ-tubulin complex proteins (GCPs) combine with the γ-TuSC to form the larger γ-tubulin ring complex (γ-TuRC); organisms lacking these additional GCPs are thought to oligomerize γ-TuSCs into similar ring complexes [12,13]. To date, only the γ-TuSC components TBG-1/γ-tubulin, GIP-1/GCP3, and GIP-2/GCP2 have been identified in C. elegans, suggesting that C. elegans γ-TuRC may share the yeast γ-TuRC composition. As our experiments do not distinguish between γ-TuSC and γ-TuRC, we will use the term γ-TuRC for simplicity. A putative C. elegans ortholog of MOZART1 (mitotic spindle-organizing protein associated with a ring of γ-tubulin, MZT1), a γ-TuRCinteracting protein and proposed γ-TuRC component, was identified based on sequence homology to the Arabidopsis ortholog, but its function has not been investigated [14]. γ-TuRC has microtubule nucleation capacity and can also cap the minus ends of microtubules, preventing minus-end growth or depolymerization [15,16]. Whether γ-TuRC predominantly functions as a nucleator or as a minus-end cap or anchor in vivo is a matter of debate.
Although γ-TuRC is essential in organisms ranging from yeast to humans, γ-TuRC depletion does not result in the elimination of all microtubules from the cell, suggesting that other mechanisms exist to grow and anchor microtubules at MTOCs. γ-TuRC removal in vivo has severely deleterious effects on mitosis, but microtubules are still present [17][18][19]. The presence of microtubules in dividing C.
elegans embryonic cells appears to rely on both γ-TuRC function and the mitotic kinase AIR-1/Aurora A [20,21], as only depletion of both TBG-1 and AIR-1 from dividing cells results in the elimination of centrosomal microtubules. Whether γ-TuRC and AIR-1, or other essential centrosomal MTOC proteins, function redundantly to build microtubules at ncMTOCs in animal cells is unknown, most notably because of the early requirement of these proteins in mitosis that prohibits an assessment of any later roles during differentiation.
In C. elegans embryonic intestinal cells, MTOC function is reassigned from the centrosome during mitosis to the apical surface as cells begin to polarize [6], thereby establishing an apical ncMTOC in each cell. Intestinal cells all derive from the 'E' blastomere, undergoing 4 rounds of division before polarizing at the 'E16' stage when the intestinal primordium is comprised of 16 epithelial cells (S1A Fig, [22]). Shortly after the E8-E16 division, the E16 cells follow a stereotypical pattern of polarization and establish their apical surfaces facing a common midline [6], the eventual site of the lumen of the epithelial tube. ncMTOCs positioned along these apical surfaces nucleate and organize microtubules into fountain-like arrays emanating away from the midline on either side [6,23,24]. Intriguingly, many centrosomal MTOC proteins including AIR-1 and GIP-1 also localize to this ncMTOC [6]. Because the earliest stages of ncMTOC formation can be easily tracked, the embryonic intestinal primordium provides an ideal system in which to test the role of specific proteins in ncMTOC establishment in vivo.
Here, we test the hypothesis that the proteins required to build microtubules at the centrosome play a similar role at ncMTOCs. To do this, we optimized an existing tissue-specific degradation strategy to test the role of GIP-1 and AIR-1 at ncMTOCs in C. elegans embryonic intestinal cells. As at the centrosome, we find that GIP-1 is required to localize the other γ-TuRC members, TBG-1 and GIP-2. Additionally, we showed that a predicted ortholog of the γ-TuRC protein MZT1 is essential in C. elegans and colocalizes with γ-TuRC in all contexts but is uniquely required for localization of γ-TuRC to the PCM, and not to the centriole or to the apical ncMTOC. This differential requirement for proteins at the centrosome versus the apical ncMTOC was a common trend, as AIR-1 was also only required to localize GIP-1 and TAC-1 to the PCM, but not to the apical ncMTOC. In addition to GIP-1 and AIR-1, we assessed the requirement of other known microtubule regulators, including ZYG-9/chTOG, PTRN-1/CAMSAP, NOCA-1/Ninein, and TPXL-1/TPX2. Surprisingly, we found that overall the depletion of these proteins did not disrupt apical microtubule organization.
Furthermore, removal of GIP-1 had only a minor effect on microtubule dynamics at the apical ncMTOC; a subset of microtubules was perturbed as indicated by a change in EBP-2/EB1 localization and dynamics. These results highlight the differences between the centrosome and other MTOCs and suggest that ncMTOCs are composed of at least two populations of microtubules, γ-TuRCdependent and γ-TuRC-independent.

Results
An optimized ZF/ZIF-1 degradation system allows for tissue-specific degradation of early essential proteins To test the role of γ-TuRC and AIR-1 at ncMTOCs, we needed a strategy to deplete proteins essential in the early embryo (early essential proteins) at later stages of development. For example, we had previously been unable to assess the function of γ-TuRC components in differentiated cells in vivo as their depletion causes severe mitotic defects that result in early embryonic lethality.
Tissue-specific degradation strategies have provided a means to deplete such early essential proteins [25,26]. We therefore optimized an existing tissuespecific degradation system (Fig 1A, B). The germline cell fate determinant PIE-1 is degraded in somatic cells in the early embryo ( [27], Fig 1C). This degradation requires a zinc finger domain 1 (ZF) on PIE-1 and the SOCS-box protein ZIF-1, which targets PIE-1 for degradation by an E3 ubiquitin ligase [27]. Previous reports found that the ZF domain could be added to any protein of interest and that protein is degraded in somatic cells by ZIF-1 ( [28], Fig 1E, G). However, endogenous ZIF-1 is only expressed in the early embryo so degradation of targets later in development requires exogenous expression of ZIF-1 [29]. The major drawback of this system is that degradation of early essential proteins by endogenous ZIF-1 leads to an early arrest. We found that a zif-1 deletion mutant is viable (92% ± 8.4% embryonic viability in zif-1(gk117) worms compared to 99% ± 1.9% in N2 worms) despite the apparent loss of ZIF-1 activity (Fig 1C, D).
Using a zif-1 mutant background ('zif-1(-)'), we can tag any gene with the ZF domain using CRISPR/Cas9 and the resulting ZF-tagged protein is not degraded ( Fig 1F, H, I, L, O), thus allowing for normal development. We then express ZIF-1 under the control of a tissue-specific promoter to degrade ZF-tagged targets.
Using this strategy, we tagged the γ-TuRC component GIP-1/GCP3, the predicted MZT1 ortholog W03G9.8 which we hereafter refer to as MZT-1 (see below), and the mitotic kinase AIR-1/Aurora A with ZF::GFP, allowing us to monitor protein expression and localization and to degrade each protein with exogenous ZIF-1 expression (Fig 1I-Q). As expected, GIP-1 localized to the apical ncMTOC in intestinal cells and AIR-1 decorated microtubules (Fig 1I, O, [6]). ZIF-1 was then expressed using the promoter for the elt-2 gene, which is expressed exclusively in the intestine starting at intestinal stage E2 (S1A Fig). ZIF-1 expression led to intestine-specific removal of GIP-1, MZT-1, or AIR-1 ('GIP-1 gut(-) ', 'MZT-1 gut(-) ', 'AIR-1 gut(-) ') as demonstrated by the loss of both apical and cytoplasmic GFP signal (Fig 1I-Q, S1B-E Fig, S2A-D' Fig). We quantified this intestine-specific depletion in two ways. First, we measured the total amount of reduction of GFP signal in the intestinal primordium of 'gut(-)' embryos as compared to 'gut(+)' siblings that lacked the ZIF-1-expressing array (% GFP depletion: GIP-1 gut(-) 93.1%, MZT-1 gut(-) 92.1%, AIR-1 gut(-) 82.1%, S3A Fig). This is likely an underestimate, due to the out of focus light contributed by non-degraded ZF::GFP in non-intestinal cells that complicates this assessment, especially when highly expressed genes like air-1 were tagged. We also took line scans across the midline of the intestinal primordium in both gut(-) and gut(+) embryos to measure apical enrichment (Fig 1K, N Methods). Thus, the zif-1 mutant coupled with the ZIF-1/ZF degradation system provides an effective tool for depleting early essential proteins in a tissue-specific manner.

Degradation of GIP-1/GCP3, MZT-1, and/or AIR-1/Aurora A in intestinal cells leads to mitotic defects, but does not impair intestinal differentiation
To test the role of γ-TuRC and AIR-1 in establishing the apical ncMTOC, we needed to effectively remove these proteins from intestinal cells prior to the E16 stage when cells reassign MTOC function to the apical membrane. Intestinal differentiation in C. elegans proceeds in the absence of cell division [30], suggesting that mitotic defects in intestinal cells per se would not affect their ability to build an ncMTOC. Thus, we could begin degradation of the desired targets during the intestinal divisions to ensure they would be effectively cleared by the time cells began to polarize and establish the apical ncMTOC.
In C. elegans, loss of maternal AIR-1 results in severe mitotic defects including multinucleate cells, polyploidy, disorganized microtubules, and failed centrosome separation [31,32]; the additional removal of γ-TuRC components results in monopolar spindles and loss of centrosomal microtubules in the first cell division [20,21]. We thus used mitotic defects in intestinal cells as a phenotypic read out for effective degradation of GIP-1 and AIR-1 prior to polarization. The E blastomere undergoes 4 rounds of division to generate the polarized 16-cell primordium (E16, Fig 2A). As ZIF-1 was expressed from an early promoter (elt-2p) that is active beginning around E2-E4, we expected that successful removal of γ-TuRC and AIR-1 should result in polarized intestinal primordia with between 2 and 16 cells. Indeed, we found that degradation of GIP-1, MZT-1, or AIR-1 resulted in embryos with 8.6 ± 2.3, 7.6 ± 1.8, or 9.2 ± 1.5 intestinal nuclei, respectively, and that degradation of both AIR-1 and GIP-1 resulted in embryos with 4.0±0.0 intestinal nuclei (Fig 2A, B). Expression of ZIF-1 from a promoter that is active around E8 (ifb-2p) also significantly reduced the number of intestinal nuclei, although to a lesser extent (Fig 2A, B, S1A Fig). As further proof that we were effectively depleting the desired targets, we found that embryos with decreased or no zygotic air-1, and with only a maternal supply of ZF::GFP-tagged AIR-1 ('AIR-1*'), had intestinal nuclear numbers indistinguishable from AIR-1 gut(-) embryos (see Materials and Methods, 9.1 ± 2.2, Fig 2B). We frequently observed mitotic defects in AIR-1 gut(-) and GIP-1 gut(-) embryos, such as scattered condensed chromosomes, binucleate cells, and abnormal mitotic spindles (S2I Fig, S1 Movie). These results are consistent with the reported role for γ-TuRC and AIR-1 in mitosis and suggest that GIP-1 and AIR-1 are effectively depleted from intestinal cells beginning at approximately E4.
We next tested whether intestinal cells can polarize and differentiate in the absence of GIP-1 or AIR-1. In intestinal cells depleted of both GIP-1 and AIR-1, the apical polarity protein PAR-3, whose localization to the apical surface is a hallmark of apico-basal polarity [33], was unperturbed (S2E- H Fig). Similarly, we observed intestine-specific lysosome-like organelles known as gut granules ( Fig   3D), which are hallmarks of intestinal differentiation [23,30]. Together, these results confirmed that we could use tissue-specific degradation to deplete GIP-1 and AIR-1 prior to apical ncMTOC formation without any dramatic effects on intestinal differentiation.   [17,39], we find that mzt-1 is also required for embryonic viability.

GIP-1/GCP3 is required to localize other γ-TuRC components to the apical ncMTOC
Our ability to deplete GIP-1 ( Fig 3G) and MZT-1 (Fig 3E)   where they appear as naked centrioles that completely lack PCM by electron microscopy [6]. At this stage, GIP-1 localizes to both the centrioles and the apical ncMTOC ( Fig 4C). By contrast, MZT-1 gut(-) embryos localized GIP-1 to their centrioles, but failed to recruit GIP-1 to the PCM during mitosis (compare Fig  PCM-specific linker for γ-TuRC in dividing cells. Consistent with these findings, human MZT1 appears to promote the targeting and activation of an intact γ-TuRC to the centrosome in human tissue culture cells [38]. In C. elegans, MZT-1 localization tracks with γ-TuRC localization at the centriole, the PCM, and the apical ncMTOC. However, MZT-1 is not required to target γ-TuRC to the apical ncMTOC but requires GIP-1 to localize there, suggesting that MZT-1 is stably associated with the complex even when not playing a targeting role.
Additionally, NOCA-1/Ninein is often found to colocalize with the minus ends of microtubules, although it has never been shown to directly bind to minus ends [44]. A previous report found that γ-TuRC and NOCA-1 function together in parallel with PTRN-1 to maintain non-centrosomal microtubule arrays in C. elegans larval and adult skin [45]. Furthermore, the NOCA-1 h-isoform appears to localize to the membrane ncMTOC in the C. elegans adult germline using a palmitoylation site [45]. In the absence of this site, γ-TuRC is required to target NOCA-1 to the ncMTOC. We found that both PTRN-1 and NOCA-1 localize to the apical ncMTOC in wild-type embryonic intestinal cells (S4A and e-isoforms lack the NOCA-1h region containing the characterized palmitoylation site; however, we did not rule out the use of alternative palmitoylation sites, which the d-isoform is predicted to contain (Cys6, [46]).

AIR-1/Aurora A is required to target TAC-1 and γ-TuRC to the centrosome in intestinal cells, but not to the apical ncMTOC
AIR-1/Aurora A is a mitotic kinase that helps activate MTOC function at the centrosome, in part by driving PCM accumulation of targets required for microtubule nucleation and polymerization, such as γ-TuRC and TAC-1/TACC [32,[47][48][49][50]. The phosphorylated, kinase-active form of AIR-1 localizes to the apical ncMTOC along with GIP-1 and TAC-1 [6], suggesting that AIR-1 could similarly regulate the accumulation of these targets at the apical ncMTOC. We first asked whether AIR-1 is required for normal GIP-1 and TAC-1 accumulation We note that centriolar GIP-1 signal appeared to remain upon AIR-1 depletion ( Fig 5B). We observed a similar reduction in GFP::TAC-1 levels at the mitotic centrosome of AIR-1 gut(-) cells compared to control cells (501.9 ± 91.7 vs. 972.0 ± 463.6, p = 2.88 × 10 -6 , Fig 5D-F), suggesting that AIR-1 can be efficiently depleted prior to ncMTOC formation. In contrast to the E8 mitotic centrosomes, GFP::GIP-1 and GFP::TAC-1 were still recruited to the apical ncMTOC in control and AIR-1 gut(-) embryos (Fig 5G-J). These results indicate that AIR-1 is required for GIP-1 and TAC-1 recruitment to intestinal mitotic centrosomes, as is known in other cell types and organisms [48,51], but that AIR-1 is not required for their localization to the apical ncMTOC, further distinguishing the centrosome from the apical ncMTOC.

Microtubules remain apically enriched upon depletion of essential microtubule nucleators, anchors, and stabilizers
γ-TuRC and AIR-1 are required to nucleate microtubules at the centrosome in C.
The surprising finding that known microtubule nucleators are not required to build the majority of microtubules at the apical ncMTOC indicates that other mechanisms or molecular players are required to perform this task. We investigated other known microtubule regulators-the anchoring protein NOCA-1/Ninein, stabilizers PTRN-1/CAMPSAP2 and TPXL-1/TPX2, and the polymerase ZYG-9/chTOG [18,52,53]-to determine if they are required to organize microtubules apically. We found that apical microtubule enrichment was slightly but significantly decreased only upon depletion of ZYG-9, but that even then, microtubules remained apically enriched (Fig 6B, D, F). We were particularly surprised to see grossly normal apical microtubule organization in [GIP-1; NOCA-1] gut(-) ; ptrn-1(0) triple mutant embryos, as noca-1 and ptrn-1 are required in parallel to maintain the organization of non-centrosomal microtubule arrays in hypodermal epithelial cells [45]. These results suggest that different MTOCs, and even different ncMTOCs, have distinct molecular and genetic requirements to generate specific microtubule arrays, and that more mechanisms remain to be identified.

A subset of apical microtubules is perturbed upon depletion of GIP-1
While we found that overall apical organization of microtubules was intact, we explored whether microtubule dynamics were altered upon depletion of microtubule regulators. One possibility is that the majority of microtubules at the apical ncMTOC are stable, persisting from the mitotic divisions prior to polarization. We tested this possibility in two ways. First, we found that upon nocodazole treatment (10 μg/mL and 30 μg/mL), apical microtubule enrichment was significantly reduced both over time and compared to control treated embryos ( Fig 6H-J), indicating that apical microtubules can be destabilized. This experiment also demonstrates that our measurement methods (line intensity profiles and apical enrichment) are sensitive enough to detect differences in varying amounts of apical microtubules.
Second, we probed microtubule dynamics at the apical ncMTOC by examining the localization of EBP-2/EB1, a microtubule plus-end-binding protein that associates with growing microtubule plus ends. To do this, we tagged the endogenous ebp-2 locus with GFP using CRISPR/Cas9 to visualize endogenous EBP-2 comets. In control embryos, EBP-2 accumulated at the apical surface, and moved along the apical surface and out along lateral microtubule tracks towards the basal part of the cell, consistent with microtubules growing from the apical ncMTOC (Fig 7A-C We next measured the speed of the EBP-2 comets coming from the apical ncMTOC to determine if the dynamics of microtubule growth were altered (Fig 7D,   F). In control embryos, apically-derived comets had an average speed of 0.558 µm/sec, which was significantly slower than the reported speeds for comets originating from centrosomes in the early embryo that had been previously labeled with overexpressed EBP-2::GFP [18] or for endogenous early embryo centrosomal comet speeds we measured ('2-cell' 0.8884 µm/sec, p = 7.68 × 10 -13 , two-tailed t-test , Fig 7D, F). We also measured comets speeds from centrosomes in the E8-E16 division ('E8', 0.563 µm/sec), which had similar speeds to apically-derived comets, and were also significantly slower than comets from centrosomes in the 2-cell embryo (Fig 7D, F), suggesting that cell type, cell size, or centrosome size may influence comet speed [54]. Surprisingly, we found that the speed of apically derived comets was significantly increased relative to controls in GIP-1 gut We demonstrate that our adapted ZIF-1/ZF degradation system is a robust method for depleting endogenous proteins in a specific tissue of interest (the primordial intestinal epithelium), thereby allowing us to probe the function of early essential genes in differentiating tissues. We monitored and characterized the effectiveness and efficiency of endogenous protein depletion by adding both GFP and ZF via CRISPR to genes of interest. Degradation of many of these critical centrosomal proteins during intestinal divisions caused mitotic defects, confirming that targeted proteins were depleted before the apical ncMTOC formed. With this adapted method now validated, future studies can omit the GFP and use ZF-tagged CRISPR alleles, which will permit a broader range of quantitative analyses of GFP markers.
A consequence of early depletion of important centrosomal proteins was fewer intestinal cells, causing architectural defects in the intestinal primordium, such as overall shorter apical midlines. However, the overall reduced midline surface, especially in [GIP-1;AIR-1] gut(-) embryos, does not explain the general upward trend in apical enrichment of α-tubulin we observed; we found no evidence of a correlation between midline length and α-tubulin enrichment (see Materials and Methods). In addition to changes in intestinal geometry, early depletion of centrosomal proteins likely also caused changes in ploidy. While these changes may have impacted zygotic gene expression, they cannot explain, for example, the observed differences among MZT-1 gut(-) , GIP-1 gut(-) , and AIR-1 gut(-) embryos, which all have similar nuclear numbers and thus likely similar ploidy defects. In sum, differences in microtubule dynamics do not appear to correlate with ploidy or architecture defect severity (Fig 2, 7), indicating that these secondary defects alone cannot account for the changes in microtubule dynamics we observe.
Using the ZIF-1/ZF system, we characterized the predicted C. elegans ortholog of MZT1, presenting the first in vivo characterization of a MZT1 ortholog in animal development to our knowledge. As in many systems, MZT-1 colocalizes with other γ-TuRC components, is required for γ-TuRC localization to the mitotic spindle pole, and is essential for viability. Surprisingly, we found that γ-TuRC does not require MZT-1 for its localization to the centrioles and apical MTOC. In addition to localizing γ-TuRC, MZT1 is important for nucleation activity of γ-TuRC in Candida and in human tissue culture cells [38,55]. However, we found that only intestinal GIP-1 depletion, and not MZT-1 depletion, impacted comet number and dynamics, suggesting that the MZT-1 found at the apical MTOC is not important for γ-TuRC activity, and may simply be a non-functional component of the γ-TuRC complex at these non-PCM sites.
Strikingly, we found that dynamic microtubules were still observed growing from the apical ncMTOC following depletion of GIP-1 and AIR-1, which are essential for centrosomal MTOC activity [20]. This finding indicates that additional mechanisms for generating dynamic microtubules must exist. One exciting possibility is that additional yet undiscovered nucleators exist in the cell.
Based on previous studies on centrosomal microtubules, these hypothetical molecules might be unique for building microtubules at ncMTOCs.
Rather than additional nucleators, another possible mechanism could be through the action of microtubule-stabilizing and -anchoring proteins, as has been seen for other ncMTOCs. In fact, the exact role of γ-TuRC in vivo is not known. The relatively poor nucleation capacity of γ-TuRC in vitro suggests that factors that activate its nucleation capacity at MTOCs exist in vivo [12,56].
Alternatively, the primary function of γ-TuRC might not be nucleation, as is suggested by imaging studies of centrosomes from γ-tubulin-depleted C. elegans embryos [57]; a large number of microtubules are still found associated with the centrosome following γ-tubulin depletion, but are disorganized relative to the centrioles. These data raise the possibility that γ-TuRC functions in anchoring microtubules onto the PCM, and that perhaps dynamic microtubules at the apical ncMTOC are generated from many different types of stabilized microtubule seeds. For example, proteins like PTRN-1/Patronin/CAMSAP and NOCA-1/Ninein could protect and anchor small microtubule seeds that grow in parallel to the γ-TuRC-based microtubules. Evidence for this type of model has been seen in Drosophila oocytes and C. elegans skin cells [45,58]. However, our analysis of GIP-1 gut(-) ; NOCA-1 gut(-) ; ptrn-1(0) triple mutants suggests that in embryonic intestinal cells, the ncMTOC does not require PTRN-1 and NOCA-1/Ninein, even in parallel with GIP-1. Furthermore, microtubules remained organized at the apical MTOC upon depletion of the microtubule polymerase ZYG-9/chTOG and the spindle assembly factor TPXL-1/TPX2, suggesting that additional microtubule regulators remain to be discovered.
A final possibility is that MTOCs could facilitate microtubule growth not by localizing nucleators, but instead by increasing the local tubulin heterodimer concentration. Microtubules can be nucleated in vitro in the absence of any additional molecules, depending on the concentration of tubulin. Recent studies of in vitro reconstituted PCM suggest that the centrosome might build microtubules in part through the selective concentration of tubulin [53]. ncMTOCs might similarly concentrate tubulin, leading to localized microtubule growth. Two of our findings are consistent with this possibility. First, we observed α-tubulin enrichment at the apical MTOC following microtubule depolymerization with nocodazole, though we cannot distinguish between free tubulin heterodimers and small protected microtubule seeds. Second, we found a small but significant decrease in apical microtubule enrichment in ZYG-9 gut(-) embryos, which can concentrate tubulin in vitro at SPD-5 condensates [53], and could perhaps play a similar role at the apical ncMTOC.
Our finding that microtubules have increased growth speeds following GIP-1 depletion suggests that a limiting growth factor is normally present at the apical ncMTOC. This limiting factor could be sequestered by γ-TuRC itself, or it could be limiting because of competition for it among the large number of growing microtubules. In the first case, we would expect loss of γ-TuRC to release this factor and cause both increased microtubule growth speed and comet number. However, we observed fewer, faster comets and decreased apical EBP-2 enrichment upon GIP-1 depletion. These observations are more consistent with a model in which loss of γ-TuRC leads to fewer growing microtubules, thereby increasing the availability of a limiting factor to growing microtubules and allowing faster growth (Fig 7G). This limiting factor could be tubulin heterodimers themselves, however, the mechanisms for concentrating a pool of tubulin at a membrane are completely unknown.
Finally, we found that γ-TuRC and AIR-1 are not required to form the majority of apical microtubules, raising the question of why these proteins so specifically localize there as a new ncMTOC is being established. One possibility is that the main function of localizing γ-TuRC and AIR-1 to the apical ncMTOC is to effectively remove them from the centrosome at the end of mitosis as centrosomal MTOC function is attenuated. We hypothesize that the ability to remove microtubules from the centrosome is an important step in mitotic exit, as hyperactive MTOC function at the centrosome has been linked to cancer [59][60][61].
Creating a sink for centrosomal microtubule regulators at an alternative site in the cell would provide a quick and effective way of maintaining the inactivation of MTOC function at the centrosome. Different cell types require a large variety of specific patterns of microtubule organization, and future work will be critical to discover the additional molecular players and mechanisms that contribute to the formation and function of different types of MTOCs.

Materials and Methods
All data used for quantitative analyses are included as S1 Data. Image files are available upon request.

C. elegans strains and maintenance
Nematodes were cultured and manipulated as previously described [62].
Experiments were performed using one-or two-day old adults. The strains used in this study are as follows:  [63,64]. Successfully edited worms were outcrossed at least 2 times before being used for subsequent experiments. sgRNA and homology arm sequences are listed in S1 Table. Embryonic viability To assess embryonic viability, 20-30 young adult hermaphrodites of each genotype were singled onto small plates and allowed to lay for 4 hours at 20°C, and then adults were removed and eggs were counted. After three days at 20°C, the number of surviving worms present on each plate was counted, and viability for each plate was calculated as the total number of L4s and adults divided by the number of eggs. When comparing N2 and zif-1(-) (JLF155) lethality, two N2 plates had more surviving worms at day three than the number of eggs initially counted, and those plates were omitted.

ZF/ZIF-1 degradation
The ZF/ZIF-1 system was executed as previously described with the following modifications. ZF::GFP tags were inserted into endogenous loci in a zif-1(gk117) mutant background using CRISPR editing techniques (see above). Exogenous  2).

Nocodazole Treatments
JLF36 embryos were treated with nocodazole as has been previously described [6]. Briefly, trypan blue-coated embryos were affixed to poly-lysine-coated coverslips and submerged in embryonic growth medium (EGM) [66] containing either 0.1% DMSO or 10 or 30 µg/mL nocodazole. E16-stage embryos with their dorsal surfaces facing the coverslip were selected for analysis. The eggshell and vitelline membrane were punctured using a Micropoint nitrogen dye laser (Andor), allowing the EGM + DMSO or nocodazole to reach the embryo. Embryos were imaged immediately following puncturing (T1, 10-45 seconds later) and then again after 10 minutes (T2).

Immunofluorescence
Embryos were fixed and stained as previously described [23]. Briefly, embryos of the appropriate stage were collected and adhered to a poly-lysine-coated slide with a Teflon spacer and covered with a coverslip. Embryos were fixed by freezecrack followed by 100% MeOH for 5 minutes. Embryos were rehydrated in PBS and incubated with rabbit α-GFP primary antibody (Abcam, 1/200) or mouse α-PAR-3 [28]  images were taken at a sampling rate of 0.5 μm. Images were processed in NIS Elements, the Fiji distribution of ImageJ ('Fiji') [67,68], or Adobe Photoshop.
Fixed images were obtained using a 60x Oil Plan Apochromat objective (NA =.1.4) on either the above system or a Nikon Ni-E compound microscope with an Andor Zyla sCMOS camera.

Analysis Considerations
We observed that different ZF::GFP-tagged alleles produced different levels of background signal in the intestinal primordium following ZIF-1-mediated depletion.
This background signal was likely due to out-of-focus light produced by the non- genotypes with AIR-1 gut(-) and genotypes with no tagged AIR-1 (Fig 7B, C). For genotypes that did not meet this quantitative cutoff, our comparisons were qualitative. These differences in background did not impede our analysis of EBP-

Quantification of GFP depletion
Pre-bean-and bean-stage embryos were used for analysis. Using Fiji, end-1p-

Quantification of centrosomal protein accumulation in AIR-1 gut(-) mutants
Despite the high background in AIR-1 gut(-) embryos (see above), we were able to measure and detect a clear difference in centrosomal TAC-1 and GIP-1 even with no correction for genetic background (AIR-1::ZF::GFP signal from outside the intestinal primordium). Embryos in the E8-E16 division were chosen for analysis. Centrosomes with strong mCherry::TBA-1 localization were considered active MTOCs and their average GFP intensity was measured. A 7-pixel diameter circular ROI was placed manually to measure the average intensity of GFP::GIP-1 and GFP::TAC-1 at the Z-section of maximum GFP intensity at active centrosomes in air-1(+) and AIR-1 gut(-) backgrounds. Signal intensities were compared using Welch two-sample t-tests.

Quantification of apical α-tubulin and EBP-2 enrichment
Pre-bean-and bean-stage embryos were chosen for analysis. Using Fiji, end-1p-

Quantification of EBP-2 comets
Time-lapse images of pre-bean and bean-stage embryos were obtained using a 100X objective (see Microscopy). All time-lapse images were processed to correct for photobleaching prior to analysis using the Bleach Correction plugin for E16 embryos with only one measured comet, images were re-analyzed to include at least two measured comets, so that each embryo's average comet speed was based on two or more measured comets. 2-cell embryos with only one measured comet were excluded from analysis. Comet speed was calculated using R. Comet count and speed were compared between genotypes using Welch two-sample t-tests. In comparing comet number between genotypes, we did not observe a significant decrease in GIP-1 gut(-) compared to control alone, but when we pooled embryos from the three GIP-1 gut (-) genotypes (G, GA, GNP) and compared them to pooled embryos with wild-type gip-1 (C, A, M), we saw a significant difference between the groups: pooled(GIP-1 gut(-) ) had an average of 7.7 ± 3.3 comets per embryo, and pooled(gip-1(+)) had an average of 11.1 ± 4.1 comets per embryo (p = 0.0001, two-tailed t-test). Importantly, none of the individual genotypes within a pooled group was significantly different from the others by paired t-tests. All of the gut(-) genotypes with EBP-2::GFP measured in Fig. 7 have excess α-tubulin added from the zif-1-expressing transgene wowEx10. To test if overexpressed α-tubulin was influencing our comet speed results, we measured comet speed in control and GIP-1 late gut(-) embryos which do not overexpress tubulin. In the absence of overexpressed α-tubulin, we still saw similar results as in Fig 7: control comet speed (0.57 ± 0.12 μm/sec) was significantly slower than in GIP-1 late gut(-) (0.69 ± 0.12 μm/sec, p = 0.008 two-tailed t-test). Figures   Fig 1. A tissue-specific degradation system to deplete early essential proteins A, B) Cartoons depicting tissue-specific protein degradation scheme (adapted from [29]). In the presence of endogenous ZIF-1 (zif-1(+)), ZF-tagged targets are degraded in somatic cells leading to an early arrest in the case of early essential proteins. zif-1(-) mutants (zif-1(gk117)) fail to degrade endogenous ZF-tagged