Divergent Contribution of the Golgi Apparatus to Microtubule Organization in Related Cell Lines

Membrane trafficking in interphase animal cells is accomplished mostly along the microtubules. Microtubules are often organized radially by the microtubule-organizing center to coordinate intracellular transport. Along with the centrosome, the Golgi often serves as a microtubule-organizing center, capable of nucleating and retaining microtubules. Recent studies revealed the role of a special subset of Golgi-derived microtubules, which facilitates vesicular traffic from this central transport hub of the cell. However, proteins essential for microtubule organization onto the Golgi might be differentially expressed in different cell lines, while many potential participants remain undiscovered. In the current work, we analyzed the involvement of the Golgi complex in microtubule organization in related cell lines. We studied two cell lines, both originating from green monkey renal epithelium, and found that they relied either on the centrosome or on the Golgi as a main microtubule-organizing center. We demonstrated that the difference in their Golgi microtubule-organizing activity was not associated with the well-studied proteins, such as CAMSAP3, CLASP2, GCC185, and GMAP210, but revealed several potential candidates involved in this process.


Introduction
Membrane trafficking along microtubules (MTs) plays an essential role in a variety of intracellular processes. Microtubule system architecture coordinates intracellular transport and determines its preferred direction and intensity. Microtubules, therefore, are often organized in a radial array that ensures efficient transport of cellular components [1]. In many cell types, the centrosome serves as a dominant microtubule-organizing center (MTOC), featuring microtubule-nucleating and -anchoring activities [2]. However, in addition to the centrosome, the spots of microtubule nucleation can also be presented on the Golgi membranes [3,4], the nuclear envelope [5,6], the cell cortex [7], and some other places, such as chromatin, kinetochores, or pre-existing MTs (reviewed in [8]). It is well known that Golgi, along with the centrosome, serves as a potent MTOC, capable of nucleating MTs and retaining them in an organized fashion. The centrosomal MTs might complement the activity of the Golgi-derived ones, which in turn ought to adjust the proper self-assembly of the Golgi [9]. In some mammalian cells, such as retinal pigment epithelium cells, RPE1, approximately half of all microtubule nucleation events are observed at the Golgi membranes [10]. To date, Golgi-derived MTs have been discovered in hepatocytes, epithelial cells, neurons, skeletal muscles, and pancreatic beta cells, where this MT subpopulation finetunes insulin secretion (reviewed in [11]).
The spatial overlap of the Golgi and the microtubule convergence center, as well as the colocalization of dispersed Golgi vesicles and individual microtubules, was observed for lines such as retinal pigment epithelium or fish melanophores. These cell lines differ in many parameters, which renders a correct comparison impossible. We chose two epithelial cell lines, Vero and BS-C-1, derived from the kidney of a green monkey (Cercopithecus aethiops), which had similar morphology. One of them, BS-C-1, had already been shown to be able to nucleate microtubules at the Golgi vesicles [24].
To understand where in the cell the microtubule network is centered, we labeled microtubules, the Golgi, and the centrosome in both the BS-C-1 and Vero cells ( Figure 1A,B). Most of the cells displayed a radial microtubule aster focused on the cell center, where the Golgi apparatus encompassed the centrosome ( Figure 1C). Microtubules could also converge to a wide central area, where the Golgi presumably had an impact on microtubule organization. Some cells possessed two independent MTOCs; in this case, the Golgi appeared as a typical cap-like shape and was placed apart from the centrosome (21.3% in BS-C-1 and 8.7% in Vero). At times, it was obvious that the microtubules were associated only with the centrosome (1.2% in BS-C-1 and 4.8% in Vero) and vice versa with the Golgi but not with the centrosome. (7.2% in BS-C-1 and 2.1% in Vero). In a small fraction of cells, the microtubule system appeared disorganized (22.2% of the Vero and in 16.6% of the BS-C-1 cells ( Figure 1C).
To summarize, our analysis showed that both MTOCs were involved in microtubule organization in both cell lines, with a slight predominance of the centrosome in the Vero cells and the Golgi in the BS-C-1 as the main MTOCs.
central transport hub of the cell.

Morphological Analysis of the Microtubule System Revealed A Difference in Microtubule Architecture between the Two Lines of Green Monkey Kidney Cells
The microtubule-organizing activity of the Golgi may vary greatly in different cells. Classical concepts about the microtubule architecture were obtained using common cell lines such as retinal pigment epithelium or fish melanophores. These cell lines differ in many parameters, which renders a correct comparison impossible. We chose two epithelial cell lines, Vero and BS-C-1, derived from the kidney of a green monkey (Cercopithecus aethiops), which had similar morphology. One of them, BS-C-1, had already been shown to be able to nucleate microtubules at the Golgi vesicles [24].
To understand where in the cell the microtubule network is centered, we labeled microtubules, the Golgi, and the centrosome in both the BS-C-1 and Vero cells ( Figure  1A,B). Most of the cells displayed a radial microtubule aster focused on the cell center, where the Golgi apparatus encompassed the centrosome ( Figure 1C). Microtubules could also converge to a wide central area, where the Golgi presumably had an impact on microtubule organization. Some cells possessed two independent MTOCs; in this case, the Golgi appeared as a typical cap-like shape and was placed apart from the centrosome (21.3% in BS-C-1 and 8.7% in Vero). At times, it was obvious that the microtubules were associated only with the centrosome (1.2% in BS-C-1 and 4.8% in Vero) and vice versa with the Golgi but not with the centrosome. (7.2% in BS-C-1 and 2.1% in Vero). In a small fraction of cells, the microtubule system appeared disorganized (22.2% of the Vero and in 16.6% of the BS-C-1 cells ( Figure 1C).
To summarize, our analysis showed that both MTOCs were involved in microtubule organization in both cell lines, with a slight predominance of the centrosome in the Vero cells and the Golgi in the BS-C-1 as the main MTOCs. Vero cells with antibodies against α-tubulin (black, inverted contrast), mannosidase II as a Golgi marker (magenta), and p150 Glued as a centrosome marker (cyan). The cell on the left demonstrates the radial array of microtubules diverging from the single MTOC, which is formed by both the centrosome and the Golgi. In the right cell, the second MTOC formed by the Golgi is placed apart from the centrosome. The outlined areas are shown enlarged in red squares. Scale bar, 15 μm. (B) Similar panel with BS-C-1 cells. (C) Analysis of MT organization in Vero and BS-C-1 cells. The number of analyzed cells is 378 for Vero and 428 for BS-C-1 in 2 independent experiments. Most cells of both lines demonstrate the radial aster of microtubules, whereas disorganized or completely chaotic microtubules were observed in only 22,2% of Vero and in 16,6% of BS-C-1 cells. When the Golgi and centrosome are spatially separated, two distinguishable centers of organization could be clearly observed (21.3% in BS-C-1 and 8.7% in Vero). Rarely, microtubules extend from the centrosome only, but not from the Golgi (1.2% in BS-C-1 and 4.8% in Vero) and vice versa, while sometimes microtubules extend from the Golgi only (7.2% in BS-C-1 and 2.1% in Vero). Vero cells with antibodies against α-tubulin (black, inverted contrast), mannosidase II as a Golgi marker (magenta), and p150 Glued as a centrosome marker (cyan). The cell on the left demonstrates the radial array of microtubules diverging from the single MTOC, which is formed by both the centrosome and the Golgi. In the right cell, the second MTOC formed by the Golgi is placed apart from the centrosome. The outlined areas are shown enlarged in red squares. Scale bar, 15 µm. (B) Similar panel with BS-C-1 cells. (C) Analysis of MT organization in Vero and BS-C-1 cells. The number of analyzed cells is 378 for Vero and 428 for BS-C-1 in 2 independent experiments. Most cells of both lines demonstrate the radial aster of microtubules, whereas disorganized or completely chaotic microtubules were observed in only 22.2% of Vero and in 16.6% of BS-C-1 cells. When the Golgi and centrosome are spatially separated, two distinguishable centers of organization could be clearly observed (21.3% in BS-C-1 and 8.7% in Vero). Rarely, microtubules extend from the centrosome only, but not from the Golgi (1.2% in BS-C-1 and 4.8% in Vero) and vice versa, while sometimes microtubules extend from the Golgi only (7.2% in BS-C-1 and 2.1% in Vero).

The Centrosome and the Golgi Have Different Impacts on Microtubule Organization in the Vero and BS-C-1 Cells
To confirm our hypothesis regarding the differential contribution of the centrosome and the Golgi to microtubule organization in Vero and BS-C-1, we disrupted each of these MTOCs separately. A direct way to disrupt the Golgi is to treat cells with Brefeldin A, which blocks vesicular transport from the endoplasmic reticulum to the Golgi, thus leading to the collapse of the latter. Generally, the microtubule system appeared more damaged in the BS-C-1 cells than in the Vero, where most of the cells retained a radial microtubule array after BFA treatment. (Figures 2A,C and S1) and even appeared more radial than in the control ( Figure 2C). Figure 2. Disruption of the Golgi or the centrosome affects microtubule organization differently in BS-C-1 and Vero cells. (A) Immunostaining of Vero and BS-C-1 cells with antibodies against α-tubulin demonstrate the microtubule architecture after Brefeldin A treatment, which causes total disruption of Golgi ribbon (mannosidase II staining is not shown). Bar, 20 µm. For multiple staining see Figure S1.
(B) Immunostaining of Vero and BS-C-1 cells with antibodies against α-tubulin demonstrate microtubule architecture after centrinone treatment, which causes loss of the centrosomes in the cells (γ-tubulin staining is not shown). Bar, 20 µm. For multiple staining see Figure S2. (C) Analysis of changes in MT organization in Vero and BS-C-1 after Golgi disruption by Brefeldin A or centrosome abolishing by centrinone B. Total number of analyzed cells is indicated near the pie plots as N. Different colors of sectors correspond to the types of microtubule organization-radial, partially disorganized, or completely chaotic. (D) Types of microtubule organization in the cells. (E) Microtubules in BS-C-1 and Vero cells and cytoplasts, containing or lacking the centrosomes. Immunostaining with antibodies against α-tubulin (white or green in enlarged squares) and γ-tubulin (yellow in enlarged squares). Vero cytoplast without the centrosome demonstrates completely chaotic microtubules. (F) Analysis of MT radiality in BS-C-1 and Vero cytoplasts, containing or lacking the centrosomes. MTs were analyzed in 23 centrosomal and 20 non-centrosomal Vero cytoplasts and 23 centrosomal and 21 non-centrosomal BS-C-1 cytoplasts.
To eliminate the centrosomes from cells, we first treated them with centrinone B, which causes reversible centriole depletion by inhibiting Polo-like kinase 4 [25]. We tested the effect of the centrinone B treatment by adding it to the culture medium at a 500 nM concentration and performed the immunostaining of cells for γ-tubulin and α-tubulin a week later (Figures 2B,C and S2). We found that the microtubule radiality degenerated to a greater extent in the Vero cells than in the BS-C-1. In fact, there were almost no Vero cells containing any sort of ordered microtubules ( Figures 2C and S2).
We then mechanically removed nuclei from the Vero and BS-C-1 cells by centrifugation and obtained a mixed population of cytoplasts containing or lacking centrosomes, which could be easily distinguished by the gamma-tubulin immunostaining. ( Figure 2E). The cytoplasts containing centrosomes obtained from both cell lines demonstrated radial microtubule aster, which looked similar to those in intact cells ( Figure 2E). The enucleated BS-C-1 cells lacking centrosomes looked similar, and their microtubules diverged from the cell center to the periphery, as was shown earlier [26,27]. Acentrosomal cytoplasts obtained from the Vero cells, in contrast, revealed a chaotic distribution of microtubules, which indicated the determining role of the centrosome in the Vero cells ( Figure 2E,F). This was consistent with the previous results showing that the relative impact of the centrosome on microtubule organization was higher in Vero cells. Thus, under centrosome depletion, the situation was the opposite of that when the integrity of the Golgi was compromised using Brefeldin A treatment. Taken together, these results confirmed our initial suggestion that the Golgi participated in microtubule organization to a higher degree in the BC-S-1 than in the Vero cells.

Different Nucleation Activity at the Golgi Membranes in the Vero and BS-C-1 Cells Coexists with the Same Activity of Their Centrosomes
We monitored the early stages of microtubule regrowth in the nocodazole washout experiments. Cells were treated with nocodazole and cooled on ice to fully depolymerize the remaining microtubules. Nocodazole was washed out for 4 min in a warm medium. The cells were then fixed and immunostained to visualize the microtubules and the Golgi ( Figure 3A). Dispersed Golgi vesicles were presented in the cytoplasm as GM-130-positive dots. In each cell, an aster formed by short microtubules extending from the centrosome was clearly visible; in addition, many short non-centrosomal microtubules arose in the cytoplasm. ( Figure 3B). We found that approximately three times more newly nucleated cytoplasmic microtubules were associated with GM-130-positive vesicles in the BS-C-1 cells than in the Vero (46.55% versus 14.68%). This allowed us to conclude that the microtubulenucleating activity of the Golgi was indeed higher in the BS-C-1 versus Vero ( Figure 3B). The activity of the alternative MTOC (Golgi) could compensate for the low activity of the primary MTOC (the centrosome) if the pool of microtubule-nucleating factors in the cell is restricted and must be redistributed between the centrosome and the Golgi [11]. To find out whether the difference in the Golgi nucleating activity was due to compensatory mech-anisms resulting from the centrosome dysfunction, we decided to compare the nucleation activity of the centrosomes in Vero and BS-C-1 cells. cleating activity was due to compensatory mechanisms resulting from the centrosome dysfunction, we decided to compare the nucleation activity of the centrosomes in Vero and BS-C-1 cells. In the nocodazole washout experiments, it was not possible to accurately quantify the number of microtubules nucleated at the centrosome, due to their high density. Therefore, we proceeded to in vivo fluorescence microscopy. We expressed GFP-fused microtubule plus-end marker EB-3 in cells; thus, we could visualize nucleation events at the centrosome and/or the Golgi by observing where the GFP-EB3 comets flew from (Movie S1 and Figure 3C) as it was performed earlier [4]. Using this approach, we counted the rate of microtubule nucleation at the centrosome (see Materials and Methods). Then, we fixed the cells and immunostained them with the antibodies to mannosidase II and dynactin subunit p150 Glued to determine the precise localization of the centrosome and the Golgi ( Figure 3E). Thus, we could accurately evaluate the nucleating activity of the centrosome by counting the EB3 comets arising from it and compare it in the Vero and BS-C-1 cells ( Figure 3D).
Analysis of the movies revealed that the nucleation activity of the centrosomes was equal in both cell lines. On average, we observed 17 microtubule nucleation events per minute at the centrosomes in the Vero cells and 18 in the BS-C-1 cells. ( Figure 3D). Thus, the hypothesis that increased microtubule-nucleating Golgi activity in the BS-C-1 cells was due to the redistribution of nucleating factors from the centrosome to the Golgi membranes was not confirmed. Therefore, we verified that in both related cell lines, the In the nocodazole washout experiments, it was not possible to accurately quantify the number of microtubules nucleated at the centrosome, due to their high density. Therefore, we proceeded to in vivo fluorescence microscopy. We expressed GFP-fused microtubule plus-end marker EB-3 in cells; thus, we could visualize nucleation events at the centrosome and/or the Golgi by observing where the GFP-EB3 comets flew from (Movie S1 and Figure 3C) as it was performed earlier [4]. Using this approach, we counted the rate of microtubule nucleation at the centrosome (see Materials and Methods). Then, we fixed the cells and immunostained them with the antibodies to mannosidase II and dynactin subunit p150 Glued to determine the precise localization of the centrosome and the Golgi ( Figure 3E). Thus, we could accurately evaluate the nucleating activity of the centrosome by counting the EB3 comets arising from it and compare it in the Vero and BS-C-1 cells ( Figure 3D).
Analysis of the movies revealed that the nucleation activity of the centrosomes was equal in both cell lines. On average, we observed 17 microtubule nucleation events per minute at the centrosomes in the Vero cells and 18 in the BS-C-1 cells. ( Figure 3D). Thus, the hypothesis that increased microtubule-nucleating Golgi activity in the BS-C-1 cells was due to the redistribution of nucleating factors from the centrosome to the Golgi membranes was not confirmed. Therefore, we verified that in both related cell lines, the centrosomes were equally active, but the Golgi membranes for some reason nucleated microtubules much better in the BS-C-1 than in the Vero.

CAMSAP3 Overexpression Alters Microtubule Organization in the Golgi-Independent Manner
Rather few (~15-20) proteins were shown to participate in the process of nucleation and the anchoring of microtubules onto the Golgi [7,8,10,11,17]. We decided to investigate which of them may be responsible for the difference in the Golgi MT-organizing activity between the Vero and BS-C-1. To assess the expression of the Golgi and microtubule-associated proteins in these cell lines, we performed a transcriptomic analysis. The experiment was repeated twice, with a high level of convergence of results. A total of 35,300 transcripts were identified in both cultures, of which 24,433 belonged to the identified genes (NCBI Sequence Read Archive project id PRJNA862476).
To our surprise, proteins previously noted as participants in the assembly of microtubules on the Golgi apparatus (GOLGA2 (GM130), AKAP9 (AKAP450), GCC2 (GCC185), CAMSAP1/2, CLASP1/2, CLASRP, CEP350, GORASP1 (GRASP65), GMAP210 (TRIP11), MTCL1, CDK5RAP1/2/3, and ninein (NIN)) were expressed in the Vero and BS-C-1 at approximately the same levels for several hundred reads each (Table 1). Myomegalins (MMG) were some exceptions. In the BS-C-1 and Vero, they were represented by the LOC103247162, LOC103224073, LOC103224075, and PDE4DIP transcripts, which were expressed at a close to zero level in both cultures and were not found in the general transcriptomic list. On the whole, the study of transcriptomes showed almost no differences in the expression level of proteins that ensure the assembly of microtubules on the Golgi apparatus in the studied cultures. CAMSAP3 was an exception as its level was 3.75 times lower in Vero than in BS-C-1 cells. However, to date, only the role of its paralog, CAMSAP2, in microtubule organization at the Golgi was demonstrated [17,20]. Nevertheless, we decided to investigate CAMSAP3, since as shown by mRNA sequencing, among the MT and Golgi-associated proteins, only CAMSAP3 was expressed differentially in the studied cell lines. Table 1. Genes related to microtubules and Golgi: Expression levels in BS-C-1 and Vero cells.

Name
Alternative Name BS-C- In the previous study, Wang with coauthors demonstrated that this protein could be important for the organization of non-centrosomal microtubules as well as for the assembly of the Golgi apparatus in a centrosome-independent manner [28]. To determine whether the level of CAMSAP3 could be the determining factor for the Golgi capacity to organize the microtubule, we decided to increase the level of CAMSAP3 in Vero cells. We transfected the Vero cells with the plasmid-encoding CAMSAP3-GFP and studied the microtubules and the Golgi. We found that even the lowest levels of expression lead to total disruption of microtubule radiality, while high levels cause microtubule bundling; the Golgi is usually fragmented in transfected cells ( Figure 4A).

TRIP11
GMAP210 1062 1366 0.78 *MMG closely related protein PDE4DIP was not found in the transcriptome list.
In the previous study, Wang with coauthors demonstrated that this protein could be important for the organization of non-centrosomal microtubules as well as for the assembly of the Golgi apparatus in a centrosome-independent manner [28]. To determine whether the level of CAMSAP3 could be the determining factor for the Golgi capacity to organize the microtubule, we decided to increase the level of CAMSAP3 in Vero cells. We transfected the Vero cells with the plasmid-encoding CAMSAP3-GFP and studied the microtubules and the Golgi. We found that even the lowest levels of expression lead to total disruption of microtubule radiality, while high levels cause microtubule bundling; the Golgi is usually fragmented in transfected cells ( Figure 4A). In the microtubule regrowth experiments, we found that, in addition to nocodazole-resistant microtubule bundles, CAMSAP3 also induced the formation of numerous non-centrosomal microtubules ( Figure 4B). However, no visible colocalization between these microtubules and dispersed Golgi vesicles was found as shown by wide-field and confocal microscopy ( Figure S3B,B'). Another intriguing observation was that CAMSAP3 overexpression exhausted the cytoplasmic pool of soluble tubulin, which was relocated to the bundled microtubules ( Figure S3A).
Thus, we could conclude that CAMSAP3 is involved in non-centrosomal microtubule nucleation, which apparently takes place independently of the Golgi membranes. As a very potent MT-nucleating protein, whose role in non-centrosomal MT nucleation was previously demonstrated [28,29], CAMSAP3 might influence the Golgi morphology indirectly through the regulation of the MT architecture.

The GCC185-CLASP2 Axis Is Not Determining for the Golgi Activity As An MTOC
Although the transcriptome analysis showed no difference in the expression level of well-known proteins that ensure the assembly of microtubules on the Golgi, such proteins could be a subject of proteasomal degradation and many other post-translational In the microtubule regrowth experiments, we found that, in addition to nocodazoleresistant microtubule bundles, CAMSAP3 also induced the formation of numerous noncentrosomal microtubules ( Figure 4B). However, no visible colocalization between these microtubules and dispersed Golgi vesicles was found as shown by wide-field and confocal microscopy ( Figure S3B,B'). Another intriguing observation was that CAMSAP3 overexpression exhausted the cytoplasmic pool of soluble tubulin, which was relocated to the bundled microtubules ( Figure S3A).
Thus, we could conclude that CAMSAP3 is involved in non-centrosomal microtubule nucleation, which apparently takes place independently of the Golgi membranes. As a very potent MT-nucleating protein, whose role in non-centrosomal MT nucleation was previously demonstrated [28,29], CAMSAP3 might influence the Golgi morphology indirectly through the regulation of the MT architecture.

The GCC185-CLASP2 Axis Is Not Determining for the Golgi Activity as an MTOC
Although the transcriptome analysis showed no difference in the expression level of well-known proteins that ensure the assembly of microtubules on the Golgi, such proteins could be a subject of proteasomal degradation and many other post-translational modifications. Thus, we decided to study the role of the best-characterized proteins such as CLASP2, whose role in MT organization on the Golgi has been supported by numerous publications (reviewed in [8,11]).
We found that the level of CLASP2 was elevated in the BS-C-1 cells compared to the Vero ( Figure 5A,B). GMAP 210 was poorly detected in cell lysates, and immunofluorescent staining revealed no morphological differences between the BS-C-1 and Vero cells ( Figure S4). Our study, therefore, was focused on CLASP2 and GCC185 as its intermediary. Localization of endogenous CLASP2 did not appear to be different between the two cell lines ( Figure 5C): The endogenous protein was distributed throughout the cytoplasm, with some affinity for the central region where the microtubules converged. The next step was to overexpress exogenous CLASP2 in the Vero cells to reawaken the Golgi microtubuleorganizing activity. GFP-fused CLASP2 overexpression in the Vero cells resulted in most of the exogenous protein being distributed along the microtubules. GFP-CLASP2 also partially co-localized with the mannosidase-II-positive vesicles in the central region of the cell ( Figure 5D). To study whether the exogenous CLASP2 could directly affect the microtubule nucleation at the Golgi membranes, we conducted microtubule regrowth experiments after the nocodazole washout in the GFP-CLASP2-transfected Vero cells. Indeed, we observed numerous short microtubules decorated with GFP-CLASP2 in the cytoplasm of transfected cells ( Figure 5E, red arrows). In addition, GFP-CLASP2 also decorated microtubules at the centrosome aster ( Figure 5E, green arrows). Visually, the transfected cells contained more cytoplasmic microtubules, which were mainly CLASP2-decorated. However, they might have remained in the cytoplasm due to the higher resistance to nocodazole, since CLASP2 strongly stabilizes microtubules. Indeed, having many short cytoplasmic microtubules was not associated with mannosidase II-positive vesicles ( Figure 5E), and the number of Golgi-associated cytoplasmic microtubules was still lower than that observed in the BS-C-1.
To enhance the binding of exogenous CLASP2 to the Golgi surface, we created a chimeric construct GFP-CLASP2-GRIP. It was previously shown that the GRIP domain was able to target GFP or even larger proteins on the surface of the Golgi apparatus [1, 30,31]. In the microtubule regrowth experiments, we observed that GFP-CLASP-GRIP did colocalize to dispersed Golgi vesicles ( Figure 5F). However, many cytoplasmic microtubules decorated with exogenous CLASP2 still originated from the foci where no mannosidase staining was observed.
Since CLASP2 does not directly bind to the Golgi membranes but uses Golgin GCC185 as an intermediary, we decided to try to express GCC185 in the Vero cells. We conducted a series of microtubule recovery experiments in the Vero cells expressing mCherry-GCC185. Exogenous GCC185, along with a homogeneous distribution throughout the cytoplasm, decorated numerous vesicles, which partially coincided with mannosidase2-positive dots. This nonetheless decreased overall levels of microtubule nucleation in transfected cells, both in the cytoplasm and at the centrosome ( Figure 5G).
These results suggest that the GCC185-CLASP2 axis is not fully responsible for the microtubule nucleation at the Golgi in our cell system. None of its components, when overexpressed, could reinforce the weak microtubule-organizing capacity of the Golgi in the Vero cells to the level of the one in the BS-C-1 cells. Eventually, one could suggest the existence of one or more unknown participants in the process of microtubule nucleation at the Golgi membranes. To explore this possibility, we decided to deeply analyze our transcriptomic data to find novel regulators of the MT organization onto the Golgi. The cells were stained with antibodies against α-tubulin (cyan) and mannosidase II (magenta). Cytoplasmic MT decorated by GFP-CLASP2 is marked with red arrows. The outlined area is shown enlarged on the right. Centrosomal astral microtubules decorated by GFP-CLASP2 are marked with green arrows. Bar, 20 µm. (F) Microtubule regrowth in Vero cells expressing chimeric construct GFP-CLASP-GRIP, which targets fused proteins to the surface of the Golgi (yellow). The cells were stained with antibodies against α-tubulin (cyan) and mannosidase II (magenta). GFP-CLASP-GRIP colocalizes to dispersed Golgi vesicles, but large proportion of cytoplasmic microtubules decorated with CLASP2 do not colocalize with mannosidase II-positive dots. Bar, 20 µm. (G) Microtubule regrowth in Vero cells expressing mCherry-GCC185 (yellow). The cells were stained with antibodies against α-tubulin (cyan) and mannosidase II (magenta). Exogenous GCC185 decorated numerous vesicles, which partially coincided with mannosidase II staining. GCC185-expressing Vero cells demonstrate decreased levels of microtubule nucleation both in the cytoplasm and at the centrosome. Bar, 20 µm.

Transcriptomic Analysis Reveals Several Hitherto Unknown Potential Factors That May Influence the Golgi Functioning as a Microtubule Organizer
We found that the transcriptome composition of the cultures differed markedly. It turned out that 63 genes were expressed two orders of magnitude higher in the BS-C than in the Vero, and 52 genes were expressed in the Vero 100 times higher than in the BS-C-1. A total of 4163 genes were predominantly (>2) expressed in the BS-C-1 than in the Vero, and 2945 genes were predominantly (>2) expressed in the Vero. Thus, approximately 29% of the identified genes differed significantly in their expression levels in the two studied green monkey kidney cultures. Hereafter, we will denote the ratio of the expression level in the BS-C-1 to the expression level in the Vero as B/V = X.
The genes expressed differentially in these cell lines belonged to very different functional groups. Consequently, we focused only on the genes related to microtubule assembly on the Golgi apparatus. Typically, most of them were assumed to have invariant expression levels.
First, we analyzed microtubule and centrosome proteins. In the Human Protein Atlas database, these proteins were represented by 1134 names, among which were microtubule structural proteins, centrosome proteins, microtubule motor proteins, and microtubule interacting proteins. We did not find 83 proteins in our transcriptomic list, and 43 genes showed zero or close to zero expression levels. Of the remaining genes, 127 were predominantly present in the BSC-1 and 91 in the Vero (Table S1). We noticed that the second isoform of gamma-tubulin (TUBG2) was highly expressed in the Vero, though this isoform usually presents in embryonic tissues. It was also shown that γ-tubulin-2 accumulates in the adult brain. γ-tubulin-2 accumulation in mature neurons and neuroblastoma cells during oxidative stress may denote its prosurvival role in neural cells [32]. The first gammatubulin isoform TUBG1 and another γ-TuRC subunit, TUBGCP, were expressed at the same level in both cultures. All other tubulins were either presented at approximately the same levels or were completely absent in both cultures.
Most surprisingly, a significant amount of MAP2, a well-known microtubule stabilizer, was present in BS-C-1 cells (B/V = 7.12) (Table S1). This culture also actively expressed the microtubule-stabilizing protein Furry FRY (B/V = 9.55), as well as the BCAS3 microtubuleassociated cell migration factor (B/V = 3.45), Tubulin polymerization-promoting protein (TPPP) (B/V = 3.5), and Doublecortin domain containing 2 (DCDC2) (B/V = 3.46). Thus, the microtubules in the BS-C-1 might be more stable than in the Vero, but this needs further investigation. In our experiments, we did not observe any difference in the sensitivity of the microtubules to nocodazole and in their growth rate upon washing. Microtubule-associated protein 1 light chain 3 gamma (MAP1LC3C) (B/V = 14.17) and Microtubule-associated protein 1 light chain 3 alpha (MAP1LC3A) were also expressed at a higher level in BS-C-1(B/V = 3.15), but both proteins were related to autophagocytosis processes rather than to microtubule regulation. In BS-C-1, the expression of the CAP-Gly domain containing linker protein family member 4 (CLIP4) was also upregulated (B/V = 3.43).
In both cultures, the centrosome proteins CEP126, CEP76, CEP295NL, and centrin-1 were not expressed or were not found in the transcriptomic data. In the BS-C-1, the expression of the centrosomal protein 85-like protein (CEP85L) was more pronounced (B/V = 3.58). Other centrosomal proteins were expressed at approximately the same levels. Microtubule stability could also be regulated by kinesins. Both cultures expressed 43 kinesin genes. KIF2B was not found in the screening results, and KIF19 was expressed at a low level only in the BS-C-1. KIF5C, KIF3C, KIF6, KIF26B, and KIF21B were upregulated in the BSC-1, while KIF5A and KIF25 were upregulated in the Vero (Table S2) In total, 919 proteins could be attributed to the Golgi apparatus (Human Protein Atlas database), of which 194 were expressed at the zero level or were absent in our transcriptome lists (Table S3). A total of 177 genes were upregulated (>2) in the BS-C-1 and 106 genes in the Vero. Among the proteins of the Golgi apparatus interacting with microtubules and more represented in the BS-C-1, we noted Janus kinase and microtubule interacting protein 2 (JAKMIP2) (B/V = 3.75), Janus kinase and microtubule interacting protein 3 (JAKMIP3) (B/V = 2.08), and Microtubule-associated monooxygenase, calponin, and LIM domain containing 2 (MICAL2) (B/V = 4.63). The protein GCC1 (GCC88) (B/V = 3.46) interacts with GCC185 and actin and participates in the formation of the trans-Golgi network and endosome transport, i.e., it could also be a regulator of microtubule assembly. FAM198b (GASK-1B) was 220 times upregulated in the BS-C-1. This protein is involved in the regulation of matrix metalloproteinase secretion, and its interaction with microtubules cannot be ruled out. The Vero cells express fewer laminins (3.29 times in total) and collagens (1.77 times) than the BS-C-1; therefore, these cells might possess less "powerful" Golgi. Therefore, they might need fewer Golgi-derived microtubules as fast tracks for cargo transport.
Thus, our comparison of the transcriptomes of the Vero and BS-C-1 cells demonstrated, on the one hand, the similarity in the level of gene expression of almost all proteins, for which their role in the organization of microtubules at the Golgi membranes was shown. At the same time, significant differences were found in the level of expression of several genes encoding poorly studied proteins that could contribute to these processes.

Discussion
In the current work, we studied how the central transport hub of the cell, the Golgi, organizes its access roads for vesicular traffic, i.e., acts as a microtubule-organizing center. We described, for the first time, a significant difference in Golgi microtubule-organizing capacity between two cell lines of apparently related origin. Specifically, we have shown that the removal of the centrosome, either mechanically or by exposure to centrinone B, had a stronger effect on the microtubule system in the Vero cells. At the same time, the BS-C-1 cells were not affected by centrosome depletion, likely due to the compensatory mechanisms provided by an alternative MTOC, the Golgi. In contrast, microtubules were more affected by the Golgi dispersion with Brefeldin A in BS-C-1. This correlates well with the previous finding, showing that the BS-C-1 cytoplasts were unable to organize microtubules in the absence of the Golgi [33] but successfully organized a microtubule aster without the centrosome [26,27]. This is also consistent with previous studies in which inhibiting microtubule nucleation at the Golgi does not affect centrosomal activity, whereas removing centrosomes modifies the microtubule nucleation activity of the Golgi apparatus [34].
The general analysis of the microtubule system in connection with the two MTOCs, as well as the ablation of both separately, provided insight into the cell preference towards one particular MTOC. However, only one of the two main features of an MTOC, the capacity to anchor microtubules, was analyzed there. To study nucleation, we conducted a series of microtubule regrowth experiments and a series of direct observations of EB3-GFP "comets" that revealed the sites where nascent microtubules emanated. We demonstrated that the centrosomes were equally active in both cell lines, whereas the nucleating activity of the Golgi was approximately three times higher in BS-C-1 cells. Taken together, these data indicate that the BS-C-1 cells possessed more "powerful" Golgi in terms of its microtubuleorganizing activity, whereas the Golgi of the Vero cells was a "weak" MTOC.
To obtain a better understanding of the mechanisms underlying the Golgi activity as an MTOC, we attempted to enhance it in the Vero cells. Since, among the MT and Golgiassociated proteins, only CAMSAP3 was expressed differentially in the studied cell lines, we studied it first. Our nocodazole washout experiments, in accordance with previously published results [28,29], showed that non-radial growth of microtubules became dominant in transfected cells. Interestingly, according to another study, the moderate overexpression of CAMSAP3 actively reduced tubulin acetylation in both epithelial and neuronal cells [35]. CAMSAP3 thus seems to serve the purpose of retaining a pool of dynamic microtubules in axons, although their microtubules are generally acetylated. In any case, no visible colocalization between multiple cytoplasmic microtubules and dispersed Golgi vesicles was found in CAMSAP3-expressing Vero cells.
Previous studies identified several key participants in microtubule organization at the Golgi. Among them, the CLASP2 protein, originally discovered as a plus-tip stabilizer [36], was of greater interest as it was shown to be an essential regulator of microtubule organization at the Golgi [4]. However, it seems obvious that such a complex process cannot be attributed to a single protein or its direct partner(s). Moreover, depletion/rescue experiments of certain proteins sometimes make it impossible to provide a clear picture of the process and list all its participants. Our next attempts to increase the Golgi microtubule-organizing activity via the expression of CLASP2 or its adapter GCC185 also failed. Therefore, we could conclude that the GCC185-CLASP2 axis did not determine the difference in the Golgi properties between the two cell lines. This gave us a reason to assume the existence of some unknown participants in the process of microtubule organization at the Golgi membranes.
Analysis of the transcriptome revealed similar expression levels for most of the known participants of MT organization onto the Golgi. Among potential candidates, one can point out the Janus kinase and microtubule-interacting proteins 2 and 3 (JAKMIP2/3) and GTPase Arl4C. JAKMIP (Janus Kinase and Microtubule Interacting Protein) is a family of three genes encoding multiple isoforms of long coiled-coil proteins [37]. JAKMIP1 (known also as Marlin-1 and GABABRBP [38]) participates in the regulation of the traffic of secretory vesicles in neuroendocrine cells [39]. It was shown that Marlin-1 binds to microtubules and Golgi membranes in neurons [40]. However, almost nothing is known about JAKMIP2 (JAMIP2) and JAKMIP3 (JAMIP3, NECC-2) functions [41]. Small GTPase Arl4c is also a protein with unknown functions, though a member of the same family, Arl4a, interacts with GCC185 in a GTP-dependent manner and is needed for the structural integrity of the Golgi [42].
Interestingly, the dominance of the centrosome as the main MTOC in the Vero cells correlates with a high level of expression of the second isoform of gamma-tubulin (TUBG2), though this isoform usually presents only in embryos and in the brain [32]. Since embryonic cells usually pass through the cell cycle much quicker than the cells from adult tissues, their centrosomes might be prone to holding their role as the only MTOC during the relatively short interphase period. It was also shown that γ-tubulin-2 could accumulate during oxidative stress in mature neurons, where the role of non-centrosomal Golgi-derived microtubules is well known.
All this suggests that complex processes such as microtubule organization cannot be easily reduced to the functioning of a few proteins. Differential expression of many genes, including the undiscovered ones, may define the state and activity of each of the MTOC. A complex interplay between the centrosome and the Golgi is still not well understood. A detailed examination of several proteins, which, according to our study, are possibly involved in the microtubule organization at the Golgi membranes, is yet to come.

Cultured Cells, Transfection, Inhibitors, Microtubule Regrowth, Cytoplast Preparation, Fixation
Cultured green monkey kidney Vero (ATCC-CCL-81) and BS-C-1 cells (ATCC-CCL-26) were taken from the lab stocks. All cells were cultured in DMEM/F12 (1:1) medium (Paneco, Moscow, Russia) supplemented with 7.5% fetal calf serum (Paneco), glutamine, and penicillin/streptomycin, at 37 • C and 5% CO 2 . The transit-LT1 reagent (Mirus Bio, Madison, WI, USA) was used for transfection according to the manufacturer's instructions. For Golgi dispersion, 3 µg/mL of brefeldin A (Sigma-Aldrich, Darmstadt, Germany) was added for 2 h. For centrosome depletion, the Plk 4 inhibitor centrinone B (Tocris, cat.# 5690) was added to the culture medium at a 500 nM final concentration, and the cells were fixed daily and immunostained to visualize the centrosomes. After 7 days of incubation with centrinone B, when no centrosomes were found, the analysis of microtubule morphology was performed. For microtubule depolymerization, 3 µg/mL nocodazole was added to the culture medium and, additionally, the cells were placed for 2 h at +4 • C. To follow microtubule regrowth, cells were briefly washed in the cold medium (+4 • C) and then transferred to the warm medium (+37 • C) for the required amount of time (2-15 min). The cytoplasts were obtained as described previously [43]. Briefly, cells were treated with nocodazole (1 µg/mL) and cytochalasin B (2 µg/mL) for 1.5 h. Coverslips were placed upside down into centrifuge tubes and centrifuged in a pre-heated Eppendorf 5414 centrifuge at 14,000 rpm for 10 min. Enucleation resulted in approximately equal numbers of cytoplasts containing or lacking the centrosome. Coverslips were washed with fresh medium and incubated at 37 • C for 2 h. Cytoplasts were then fixed and proceeded to immunofluorescence.

Immunofluorescent Staining and Microscopy, Live Imaging and Data Analysis
Cells were briefly washed with PBS and fixed. Standard fixation protocol comprised two steps: Pre-fixation in absolute methanol for 7 min at −20 • C followed by subsequent fixation with 3% paraformaldehyde for 20 min at 4 • C. For immunostaining, primary antibodies were used at a concentration of 1-5 µg/mL, which normally corresponds to a dilution of 1:200 from their initial commercial concentrations. Secondary antibodies were used at the concentration of 5 µg/mL. The incubation time with both primary and secondary antibodies was 1 h at room temperature in the dark in a wet chamber. Coverslips were mounted in Aqua PolyMount (Polysciences, PA, USA) Images were acquired using the Carl Zeiss Axiovert 200 M microscope supplied with 12-bit Axio-CamHR and 12-bit AxioCamMRc cameras and AxioVision software (Carl Zeiss, Jena, Germany). For live microscopy, cells were maintained at 37 • C. The temperature was maintained by the heating unit Tempcontrol 37-2 Digital (PeCon GmbH, Bielefeld, Germany) supplied with an incubation system for live observation. Alternatively, an inverted Olympus IX71 was used accompanied by a 14-bit CCD-camera Olympus XM10 and shutter UniBlitz D122 under the control of Micro-Manager software. Cells with a relatively low level of overexpression were selected for time-lapse microscopy. During live cell imaging, frames were acquired with an interval of 2 s. Frame-by-frame viewing of the image stack allowed us to manually mark and calculate the EB3 comets originating on the centrosome, since the centrosome was also clearly seen as a bright dot. The appearance of an EB3 comet from the point marking the centrosome was scored as a microtubule nucleation event. The average number of events per minute was analyzed as described previously [44]. To analyze MT nucleation on the Golgi during nocodazole washout, the total number of short microtubules that arose in the cytoplasm was counted in each of the analyzed cells. Then it was found how many of them co-localized with GM130-positive vesicles. Thus, for each of the 25 Vero and 25 BS-C-1 cells, the percentage of cytoplasmic microtubules extending from the Golgi vesicles was calculated from the total number of microtubules in the cytoplasm. The data were analyzed using ImageJ (National Institutes of Health, Bethesda, MD, USA), Excel (Microsoft, Redmond, WA, USA) and Origin 6.1 (OriginLab Corporation, Northampton, MA, USA) software.
For confocal imaging, a Zeiss LSM900 confocal microscope was used (provided by the Moscow State University Development Program).