 Research article
 Open access
 Published:
Structure modeling hints at a granular organization of the Golgi ribbon
BMC Biology volume 20, Article number: 111 (2022)
Abstract
Background
In vertebrate cells, the Golgi functional subunits, ministacks, are linked into a tridimensional network. How this “ribbon” architecture relates to Golgi functions remains unclear. Are all connections between ministacks equal? Is the local structure of the ribbon of functional importance? These are difficult questions to address, without a quantifiable readout of the output of ribbonembedded ministacks. Endothelial cells produce secretory granules, the WeibelPalade bodies (WPB), whose von Willebrand Factor (VWF) cargo is central to hemostasis. The Golgi apparatus controls WPB size at both ministack and ribbon levels. Ministack dimensions delimit the size of VWF "boluses” whilst the ribbon architecture allows their linear copackaging, thereby generating WPBs of different lengths. This Golgi/WPB size relationship suits mathematical analysis.
Results
WPB lengths were quantized as multiples of the bolus size and mathematical modeling simulated the effects of different Golgi ribbon organizations on WPB size, to be compared with the ground truth of experimental data. An initial simple model, with the Golgi as a single long ribbon composed of linearly interlinked ministacks, was refined to a collection of miniribbons and then to a mixture of ministack dimers plus long ribbon segments. Complementing these models with cell culture experiments led to novel findings. Firstly, onebolus sized WPBs are secreted faster than larger secretory granules. Secondly, microtubule depolymerization unlinks the Golgi into equal proportions of ministack monomers and dimers. Kinetics of binding/unbinding of ministack monomers underpinning the presence of stable dimers was then simulated. Assuming that stable ministack dimers and monomers persist within the ribbon resulted in a final model that predicts a “breathing” arrangement of the Golgi, where monomer and dimer ministacks within longer structures undergo continuous linking/unlinking, consistent with experimentally observed WPB size distributions.
Conclusions
Hypothetical Golgi organizations were validated against a quantifiable secretory output. The bestfitting Golgi model, accounting for stable ministack dimers, is consistent with a highly dynamic ribbon structure, capable of rapid rearrangement. Our modeling exercise therefore predicts that at the finegrained level the Golgi ribbon is more complex than generally thought. Future experiments will confirm whether such a ribbon organization is endothelialspecific or a general feature of vertebrate cells.
Background
In multicellular organisms, such as plants and most animals, the Golgi apparatus is a multicopy organelle with subunits, the ministacks, scattered throughout the cell, often in close contact with endoplasmic reticulum exit sites [1,2,3]. In contrast, in vertebrates ministacks link to each other into a tridimensional network, the Golgi ribbon [4,5,6]. This Golgi arrangement is dynamic, undergoing regulated disassembly/reassembly during mitosis, directed migration and polarized secretion [6]. The ribbon also disassembles in response to membrane depolarization, raised intracellular calcium, viral infection, autophagy induction and genotoxic insults [7,8,9,10,11]. The importance of the Golgi ribbon in the physiology of vertebrate cells is further underscored by accumulating observations showing that a host of human diseases, especially neurodegenerative, exhibit disruption of ribbon morphology, collectively referred to as “Golgi fragmentation” [12,13,14]. Nevertheless, the biological functions of the Golgi ribbon remain unclear [15], likely for two main reasons. First, understanding of the contributions of individual components of the complex molecular machinery involved in ribbon formation and maintenance is incomplete [16,17,18,19,20,21,22,23,24,25,26,27,28]. Second, a lack of cellular systems where the detailed structure of the ribbon is intimately linked to a quantifiable biological function has prevented much experimental probing of the effects of this Golgi organization on a measurable biological output. Recently, this latter situation changed. WeibelPalade bodies (WPBs) are endothelialspecific uniquely rodlike secretory granules with lengths varying over a 10fold range (0.55μm). WPB length reflects the duallevel structural arrangement of the vertebrate Golgi apparatus. The major WPB cargo, the prohemostatic secretory protein Von Willebrand Factor (VWF), is molded into boluses during its cistrans passage through the functional subunits of the Golgi apparatus, the ministacks. Given the narrow size range of the ministacks, these VWF boluses also occur within a similar small range [29]. At the Golgi exit site, the continuous structure of the endothelial transGolgi network (TGN) allows copackaging of adjacent VWF boluses into a linear array that drives the formation of nascent WPBs in an AP1 complexdependent process [29,30,31]. The ribbon architecture of the Golgi apparatus thus determines the size of WPBs, which reflects the number of boluses copackaged at biogenesis (Fig. 1A). For this reason, these VWF boluses were dubbed “quanta” [29].
WPB size determination can be summarized thus:

i)
Ministack dimensions determine the size of VWF quanta. Changing the size of the ministacks also changes VWF quantum size and, as a consequence, WPB length.

ii)
At the TGN, which in endothelial cells forms a continuous compartment spanning multiple ministacks, the ribbon configuration allows for VWF quanta occupying adjacent ministacks to be copackaged into the same secretory granule. Unlinking of the ribbon, which also affects the TGN, results in WPB size distributions shifting to shorter lengths.

iii)
The level of VWF expression controls the fraction of ministacks occupied by quanta; i.e., the number of VWF quanta formed. As a consequence, reducing VWF expression lowers the number of quanta within a Golgi ribbon, though not their size, resulting in fewer and shorter WPBs.
Statements iiii are based on experimental evidence [29, 32].
This model of secretory granule biogenesis therefore invokes a key role for Golgi architecture, against which the level of VWF cargo controls both number and size of WPBs. Based on these premises, it is possible to mathematically model the effects of different organizations of the Golgi ribbon on WPB size and compare such simulations with experimental data obtained using a morphometric analytical workflow we developed to count numbers and measure the size of WPBs [29]. The aim of the present work was therefore to exploit the quantifiable readout of WPB size distribution to provide insight into possible ministack arrangements within the Golgi ribbon, which future experimental work might validate or disprove. The bestfitting model obtained with this approach indicates that the WPB size distribution observed in endothelial cells could be generated by a Golgi organization in which the ribbon physically integrates ministack monomers and dimers. Therefore, mathematical modeling of a system where the relationship between structure and function of the Golgi ribbon is clear and quantifiable raises the intriguing possibility that vertebrate ministack subunits are structurally  and presumably functionally  different.
Results
The experimental data against which models were tested were generated with automated highthroughput morphometric analyses of cultured primary human umbilical vein endothelial cells (HUVECs). Immunostained WPBs were segmented to extract their length (Fig. 1B). Our model of WPB biogenesis (statements iiii, above) is based on the central role of the ministack size in constraining the size of VWF quanta. We have shown that not only ministack and quantum size are similar but, importantly, changing the dimensions of ministacks impacts both VWF quantum and WPB sizes [29]. In control conditions, the median diameter of a Golgi ministack in HUVECs as measured in electron micrographs is 0.612 μm, which correlates with the median length of a VWF quantum (0.576 μm), determined by superresolution microscopy [29]. In order to simplify mathematical modeling, we equated ministack length (I) to the median size of a VWF quantum and expressed WPB lengths in multiples of l (by rounding up to the nearest integer), thus transforming a distribution of WPB size from length (Additional file 1: Fig. S1A) to number of quanta (Q), or size class (i.e., 1QWPB, 2QWPB, etc.; Fig. 1C).
Modeling the Golgi ribbon
To model how the Golgi generates this observed WPB size distribution, we started with the simplest assumptions (A):

A1: the Golgi apparatus is a single, very long (effectively infinite) ribbon composed of linearly interlinked, identical ministacks, each of fixed length l^{Footnote 1}.

A2: individual ministacks are occupied at random with a probability p by a VWF quantum.

A3: VWF quanta occupying adjacent ministacks will be packaged together into the same WPB. If, for example, five contiguous ministacks are occupied, with empty ministacks at both sides, a WPB of length five quanta (a 5QWPB) will form at the TGN.
The Golgi ribbon as a ministack linear array
Given the assumptions above, we can then model the ribbon as a “ministack linear array”, by denoting with N the number of quanta in a WPB (i.e., its size) and with p the probability of VWF quanta occupancy of ministacks across the Golgi ribbon, allowing calculation of the expected distributions of WPB lengths when p varies (Fig. 1D) with the following equation.
Independently of the value of p, all WPB length distributions predicted with this model display a mode at N = 1 (Fig. 1D) and therefore do not simulate the experimental data that show a mode at N = 2 (Fig. 1C). Furthermore, the 2QWPB frequency in the experimental distribution, 0.366, is much higher than expected by the modeling at any value of p; in fact, the greatest proportion of 2QWPBs predicted by the model is 0.25 for p = 0.5 (Fig. 1D, middle panel).
Variable ministack occupancy
We next considered whether the level of ministack occupancy might affect the relative frequency of the WPB size classes. Assuming that the rates at which WPBs are formed and that the rates at which they disappear from the cells, either because of exocytosis or degradation, are sizeindependent, one would expect cells with ministack occupancies close to 0.5 to have the most WPBs, since at very low occupancy there should be fewer WPBs, most of which should be 1QWPB, whereas at very high occupancy there should also be few WPBs but they should be close to the length of the Golgi ribbon.
Since celltocell variability in the levels of VWF synthesis is observed in endothelial cells [33] and the HUVECs used in our lab are pooled from multiple donors, the probability of occupancy of Golgi ministacks by VWF quanta varies from cell to cell (Fig. 1B and Additional file 1: Fig. S1B). Trying to explain the observed WPB size distribution, captured from a cell population, by assuming a single value of p may therefore be an oversimplification. We thus modified the “the ministack linear array” predictive model by posing that p follows a probability distribution. If p is uniformly distributed between 0 and 1 within a population of endothelial cells, we obtain:
For every value of p this model predicts that the proportion of 2QWPBs is always ≤ 0.25 and the proportion of WPBs of length N strictly decreases with N. Therefore, even accounting for VWF expression variability does not explain the high proportion of measured 2QWPBs (Fig. 1E).
Partial conclusion 1
Our initial model is based on assumptions A1, A2 and A3. Assumptions A2 and A3 derive from our model of WPB size determination (statements iii and ii, respectively), which are experimentally supported [29]. Assumption A1, a very long, single ribbon formed by a “linear array of ministacks”, is therefore likely to be incorrect, failing to generate the observed excess of 2QWPBs when modeled (Fig. 1D and E).
The Golgi as a collection of miniribbons
Alternatively, the ribbon might be formed from a collection of shorter, independent Golgi segments, ministack pairs included, which could act as preferential sites of formation for 2QWPBs. If p is sufficiently large, more 2QWPBs than 1QWPBs will form from paired ministacks; whereas at the lowest values of p only 1QWPBs will form. We tested whether a ribbon arrangement formed of short segments, a “miniribbon collection” model, might explain the observed frequency of 2QWPBs.
We modeled a Golgi made of a collection of miniribbons formed by elements of l = 3 (shown schematically in Fig. 2A) and l = 4, which give rise to WPBs of size 1Q3Q and 1Q4Q, respectively, with the frequencies shown in Table 1.
In the first case, the maximum proportion of 2QWPBs predicted is 0.268 (for p = 0.75, Fig. 2B), while in the second it is always < 0.3 (Additional file 2: Fig. S2A).
Partial conclusion 2
A Golgi "ribbon" arranged in segments of linked but independent ministack trimers and tetramers does not explain the observed 2QWPB frequency. Moreover, such a “miniribbon collection” model predicts once again that the frequency of 1QWPBs should exceed that of 2QWPBs at any value of p, except 1 (Fig. 2B and Additional file 2: Fig. S2A), and thus do not correctly simulate the experimental data.
From these simple modeling exercises it appears that the only way to get a frequency of 2QWPBs exceeding 0.3 is to have a significant number of functionally paired ministacks. Ultimately, if p = 1, then having a Golgi ribbon composed of ministacks linked into functional units whose length distribution is equal to the WPB length distribution will obviously match the data. However, the abovementioned celltocell variability in VWF expression implies that at the cell population level the probability of ministack occupancy, p, cannot be 1.
The “free ministack dimers plus linear ribbon” model
Let us then assume that the Golgi ribbon consists of a mixture of ministacks arranged into functional dimers plus long ribbon segments (depicted in Additional file 2: Fig. S2B and Fig. 2C). Equation 2 describes this “free ministack dimers plus ribbon” model, where q represents the proportion of ministacks existing as dimers.
In conditions where occupancy probability, p, is > 0.5 and q is sufficiently high, this model does predict an excess of 2QWPBs compared to 1QWPBs (Fig. 2D and Additional file 2: Fig. S2C). However, the “free ministack dimers plus linear ribbon” model predicts a proportion of 1Q and 3QWPBs higher and lower, respectively, than that experimentally measured (Fig. 1C).
Partial conclusion 3
While the “free ministack dimers plus linear ribbon” model of Golgi organization might account for the high proportion of 2QWPBs in the experimental data, it does not account for the proportion of 1Q and 3QWPB.
1QWPB instability
A major assumption built into our modeling so far is that WPBs of different sizes behave similarly; meaning that they are generated, stored and disappear from cells with identical kinetics. However, the unexpectedly low frequency of 1QWPBs might be due to their being less stable than larger WPBs. We modeled this possibility by assuming that for every currently forming WPB of length greater than one, there are β times as many WPBs of that length in the cell, whereas for 1QWPBs, there are β’ times as many. We can therefore derive a value, α:
where α describes the relative instability of 1QWPBs.
In these conditions, WPB frequency is given by:
We can fit the value of p for this distribution to the experimental data by minimizing the KolmogorovSmirnoff (KS) distance, that is the difference between the theoretical and experimental distributions for n ≥ 2. We find a bestfit value of p of 0.5692.
Using Eq. 3 for n = 1, we can then fix α to give us the correct proportion of 1QWPBs:
For these values of p and α, the modelpredicted and the experimental distributions do indeed show excellent agreement (Fig. 3A).
These results prompted us to further examine the possibility of 1QWPBs’ instability. If 1QWPBs are preferentially lost after biogenesis, they might be present in higher proportions where they are assembled; i.e. near and at the Golgi. We imaged and separately quantified the size of WPBs present in the vicinity of the Golgi versus in the rest of the cytoplasm. The periGolgi pool of WPBs indeed shows a higher fraction of 1QWPBs compared to those imaged in the cell periphery (Fig. 3B and C). These data raise the question of potential mechanisms underlying the differential depletion of the 1QWPBs. Endothelial cells control which sizes of WPBs are selected for exocytosis [34]. Therefore, one possibility is that 1QWPBs disappear by exocytosis. Release of VWF stored in WPBs occurs either acutely following exposure to agonist (stimulated exocytosis) or tonically, in the absence of any stimulus (basal exocytosis) [35]. WPBs are anchored to actin fibers adjacent to the cell membrane via the effectors Rab27 and MyRIP. Ablation of either of these proteins increases the release of VWF through basal exocytosis [36]. Does basal release differentially affect WPBs of different size classes? Depleting cells of MyRIP increases the fraction of long WPBs (Additional file 3: Fig. S3A and S3B), changing WPB size distribution (Fig. 3D). The reduced proportions of 1QWPBs indicate that secretory granules of this size class are preferentially depleted by basal exocytosis, and thus more readily disappear from the cell than longer organelles. The relative periGolgi enrichment of 1QWPBs and their differential propensity to undergo basal exocytosis thus may, at least partly, explain the observed secretory granule size distribution.
Partial conclusion 4
1QWPBs are more prone to basal exocytosis than longer WPBs.
Stable ministack dimers
Our WPB biogenesis model posits that organelle length is determined by VWF quantum occupancy of adjacent ministacks (statement iii). The “free ministack dimers plus a long ribbon” model tested earlier, while failing to explain the frequencies of 1Q and 3QWPBs, however suggested that a large proportion of ministack dimers within the ribbon might explain the measured frequency of 2QWPBs. We therefore explored whether we could detect the occurrence of ministack dimers in our system. Microtubules, their regulators and dynein motors are necessary for ministack tethering and Golgi ribbon formation [18, 22, 23]. Nocodazole depolymerizes microtubules, leading to ribbons unlinking into ministacks [22, 37]. We thus set out to quantify the effects of nocodazole treatment on WPB size distribution. Exposure to the strong secretagogue PMA (phorbol 12myristate, 13acetate) clears cells of preexisting WPBs (Additional file 4: Fig. S4A) [38]. We implemented this treatment followed by recovery in the presence of nocodazole for 24h (Additional file 4: Fig. S4B). Compared to that of DMSOpretreated/nocodazolechased cells, the WPB size distribution of PMApretreated and nocodazolechased cells shows a dramatic fall in the WPB classes > 2Q, compensated by a marked increase in 1QWPBs; remarkably, the proportion of 2QWPBs was unchanged (Additional file 4: Fig. S4C). Golgi fragments generated by nocodazole treatment are scattered throughout the cell (Additional file 4: Fig. S4B), facilitating their morphometric quantification. Golgi elements were thus measured and rounded up to the nearest multiple of l, as done for WPBs. This analysis revealed that Golgi elements were almost exclusively of sizes 1l and 2l, closely matching the distribution of WPB size classes (Fig. 4A). These data are consistent with repeated observations that the Golgi ribbon is necessary to the formation of long WPBs [29, 32, 39]; and, most importantly, they strongly suggest that even in the absence of microtubules, the Golgi of endothelial cells is not reduced exclusively to individual ministacks, but that around 50% of the Golgi objects generated by nocodazole treatment are of a size consistent with ministack dimers.
Inspection of the ultrastructure of the Golgi apparatus indeed showed the frequent occurrence of paired ministacks following nocodazole treatment, indicating they are stable structures in the absence of microtubules (Fig. 4B and Additional file 4, Fig. S4D).
Partial conclusion 5
In the absence of microtubules, the endothelial Golgi apparatus disassembles into monomer and stable dimer ministacks.
Previously, we have shown that reduction in VWF levels is reflected in the production of shorter and fewer WPBs (see statement iii regarding WPB size determination and reference [29]). Based on these observations, if ministack monomers and dimers were to persist in the ribbon, lowering VWF levels should increase the frequency of 1QWPBs and possibly affect the frequency of 2QWPBs less than that of longer organelles. We tested this expectation by treating HUVECs with different amounts of VWFtargeting siRNA to gradually reduce VWF cellular levels and therefore p (Fig. 4C). As expected, we find that 1QWPB frequency increases with the reduction of VWF levels and, strikingly, even when VWF levels are at their lowest (Fig. 4C, siVWF: 200 pmol), the proportion of 2QWPBs remains high (blue bars), an indication that ministack dimers may be a preferred unit of WPB assembly. For each condition, we also modeled WPB frequencies (red stars) as done in Fig. 3A. We note that the model fits are not as good as in Fig. 3A, because we have required α to be constant across all four conditions (see Methods for details), where it thus has to explain more data. Allowing α to vary between the conditions results in much better fits^{Footnote 2} but we could not justify this assumption biologically. These results are consistent with the possibility that the excess of 2QWPBs reflects the presence of functionally different ministack dimers within the ribbon, which would therefore be assembled from ministack monomers and these stable dimers. From this, it follows, again, that assumption A1 (ministacks are identical) is not likely to be true. VWF quantum coassembly during WPB formation may therefore not be a purely random process, which would require modification of assumption A2 in its present formulation.
Partial conclusion 6
Reduction of cellular VWF does not impact 2QWPB frequency as much as that of longer organelles, indicating that stable dimers may persist in the assembled ribbon and provide a preferred site for WPB formation.
The ribbon as a mixture of ministack monomers and dimers
Based on the findings above, we tried to assess how strong the preference for the ministack dimer vs monomer configurations might be by modeling binding rates of Golgi elements. We assumed that in the presence of intact microtubules, the Golgi elements continuously undergo a linking/unlinking process (“breathing”); that the rate at which two elements join is independent of their size; and that unlinking is slower for ministack dimers.
Let the binding rate be r, and the unbinding rate d_{1} if it does not split a stable dimer, and d_{2} < d_{1} if it does. To simplify the modeling, we assume that two neighboring monomers always form a dimer, so that in longer sections of the ribbon, dimers can be separated, at most, by one monomer. We note that tetramers can have two forms either consisting of two coupled dimers or a dimer with a monomer on either side. To make the system of equations finite (see below), we also ignore portions of ribbon with lengths > five. We let the number of portions of length n be l_{n}, for n = 1 to 5, and let M be the total number of ministacks. We denote by l_{4}, the number of tetramers consisting of two dimers and l_{4}’ the number of tetramers consisting of a dimer and two monomers. There are two types of pentamers: each has two dimers and one monomer, but the monomer can be on the end or in the middle. These need to be considered separately, since they have different unbinding rates to the two types of tetramer (Additional file 5: Fig. S5). We denote the number of pentamers with a monomer on the end by l_{5} and the number with a monomer in the middle by l’_{5}. We thus obtain a reaction system (Additional file 6: A). We can also represent the evolution of the mean numbers of each type of Golgi piece by differential equations, see Eq 4 (Additional file 6: B). The final values of the numbers of Golgi fragments of different lengths are given by the steady state values of these equations. Depending on the values of the parameters r, d_{1}, d_{2} and the total number of ministacks, the distributions can peak at dimers and can have trimers either more or less frequent than tetramers. This is illustrated by stochastic simulations of the system of reactions, changing the value of the unbinding rate d (Matlab code provided in Additional file 7). Two examples are shown: in Fig. 5A, pieces of length three are more abundant than those of length four in one simulation, whereas in the other the opposite is true.
Whether tighter linking of dimers leads to an abundance of Golgi pieces of length four (and potentially all multiples of two) therefore depends on the parameters of the system. We now fit the data. We are interested in the steady state frequencies of Golgi pieces of different length; therefore, in the model, we are free to choose both the total number of Golgi pieces (although this should be large to accurately sample the predicted frequencies) and the rate at which the system reaches equilibrium. The only combinations of parameters that should affect the steady state frequencies are rM/d_{1} and d_{1}/d_{2}. We therefore only need to explore different values of these two dimensionless parameter combinations. We can do so by arbitrarily fixing M (subject to it being large enough to get good sampling) and r and varying d_{1} and d_{2}. Therefore, we fix r = 0.001 and M = 100,000 (modeling a population of cells), both of these are dimensionless, and fit d_{1} and d_{2} to the data on proportions of Golgi of different lengths.
We show in Fig. 5B the distributions of lengths of Golgi fragments from two actual experiments, in which cells were treated with nocodazole, and from the model fit to those experiments. The reason we used these data is that in the control conditions, i.e. in the context of a linked ribbon, individual Golgi elements cannot be accurately measured, while this is possible after nocodazole treatment (Fig. 4A and Additional file 4: Fig. S4B). These simulations show a good fit to the experimental data. The calculated values of the dissociation rates based on the experimental data are d_{1} = 333 and 299 and d_{2} = 13 and 11 (chosen to minimize KS distance), respectively. Thus, the prevalence of dimers may be explained by a striking 30fold difference in ministack dissociation rates. We note that the actual values of d_{1} and d_{2} depend on our choice of the dimensional time corresponding to a unit of dimensionless time, which we cannot estimate from the steady state proportions of Golgi lengths.
Having approximately 50% of the Golgi ribbon composed of (possibly 30fold) moretightly linked dimers might explain why nocodazole fails to unlink the Golgi entirely into monomers (Fig. 4AB and schematized in Additional file 4: Fig. S4D). Unlike the “free ministack dimers plus ribbon” model (Fig. 2CD and Additional file 2: Fig. S2BS2C), where we considered ministack dimers as existing separate from the ribbon, by having 50% of Golgi objects as functional dimeric ministacks that are incorporated into the ribbon (Fig. 5C), it is possible to assemble longer organelles on these.
Partial conclusion 7
Modeling ministack binding finds that considerably different rates of dissociation between ministacks might account for the observed dimer stability even in the absence of microtubules.
Discussion
WeibelPalade bodies may represent a unique cellular system in organelle sizecontrol due to the twotiered functioning of the Golgi apparatus in their biogenesis. Golgi ministack dimensions constrain the size of VWF quanta, while the Golgi ribbon architecture controls copackaging of VWF quanta into nascent secretory granules [29]. As a consequence, the structural arrangement of a biosynthetic compartment, the Golgi, is closely reflected by the size of the organelles it generates, the WPBs. The size distribution of WPBs thus provides a proxy for the architecture of the Golgi apparatus. Taking advantage of this unique cellular system we generated a formal description of the structural organization of the Golgi apparatus. Hypotheses about the Golgi ribbon structure, translated into mathematical modeling, predicted WPB size distributions that were tested against those actually measured using an image analysis workflow that we previously employed in a series of studies [29, 32, 34, 39,40,41,42].
The measured size distribution of WPBs posed two challenges to modeling of the Golgi ribbon structure. First, why are 1QWPB depleted with respect to other size classes? Second, what kind of ministack arrangement can explain the high frequency of 2QWPBs with respect to other size classes?
For the Golgi ribbon structure to be consistent both with these data, plus what is known of WPB biogenesis, its model had to be refined from that of a single very long ribbon made by linearly linked ministacks, through to that of an ensemble of short ribbon stretches or a combination of ministack dimers and a ribbon, finally to an organization involving a mixture of ministack monomers and dimers.
For 1QWPBs, mathematical modeling suggested their low frequency might be explained by a selectively shorter residence time. Experimental testing of this hypothesis indeed indicates that 1QWPBs undergo basal exocytosis at higher rates than longer WPBs. In agreement with this finding, we had previously found that short WPBs are more prone to exocytosis [29, 32]. Of note, the modeling exercise deployed in this study allowed us to narrow down this effect to 1QWPBs, as those preferentially selected for basal secretion.
Regarding the second question, the observed steadystate frequency of 2QWPBs can, in principle, be explained by assuming that the Golgi ribbon contains a large proportion of ministack dimers. This prompted experimental testing combining treatment with PMA and nocodazole, which deplete the cellular pool of WPBs and break up the ribbon into ministacks, respectively. Strikingly, in these conditions we observed that 1Q and 2QWPBs were almost exclusively made, with their size matched by that of Golgi elements. Since there is no experimental evidence from the literature for quantization of ministack size, we had to conclude that microtubule disruption generates Golgi elements existing as ministack monomers and dimers. This conclusion was confirmed by electron microscopy observations. Thus, a structural basis must exist to prevent ministack dimer dissociation. Based on the ratio of ministack monomers and dimers following nocodazole treatment, we ran simulations that resulted in an estimated “strength” of the link between ministacks in a dimer being approximately 30fold greater than that of monomers with dimers in the context of a ribbon (Fig. 5 and Additional files 5, 6 and 7). The modeling of our experimental data thus fits with a picture of the endothelial Golgi ribbon as a dynamic structure formed by ministack monomers and very stable dimers, accounting for the measured frequency of 2QWPBs.
Intriguingly, actindependent paired ministacks were described in insect cells, displaying a Golgi organization of scattered elements, typical of invertebrate animals [43]. Following actindepolymerizing drug treatment, Kondylis and coworkers found that the number of Golgi elements increased by 50%, as would be expected if, similar to nocodazoletreated HUVECs, only half of the ministacks were dimers (refer to Fig. 2B in reference [43]). Moreover, in HUVECs individual ministack incorporated into dimers tend to sit at an angle with respect to each other (Fig. 4B); a similar configuration to that observed in insect ministack pairs (see Fig. 5AB in reference [43]). These authors further showed that in mammalian HeLa cells, pretreated with nocodazole to unlink the ribbon, inhibition of actin polymerization caused Golgi elements’ scission, suggesting the likely presence of ministack pairs also in mammalian cells [43]. One can speculate that ministack dimers may be a widespread feature of the Golgi apparatus of animal cells.
While our modeling exercise and the experimental manipulations thereby prompted support the presence of stable, microtubuleindependent ministack dimers in mammalian endothelial cells, at this stage it is not clear how these dimers are kept together. Based on the abovementioned work [43], the actin cytoskeleton is the likely driver of stable ministack dimer formation. This is consistent with the evidence that either actin depolymerization or hyperpolymerization at the Golgi results in ribbon fragmentation [24, 25]. However, precisely which of its regulators are involved remains to be experimentally established. Golgi membranes certainly provide a scaffold for many candidates [44].
Adopting WPB size distribution as a proxy for modeling the Golgi ribbon organization thus paints an unanticipated picture of this architecture. Our bestfitting model is consistent with a ribbon made from similar numbers of ministack monomers and dimers undergoing continuous linking/unlinking: a “breathing” ribbon model (Fig. 5C). This “breathing” would make the Golgi ribbon a highly dynamic structure, capable of promptly responding to signals that require its rearrangement, such as repositioning and reorientation during migration and directed secretion [23, 45, 46]. It is possible that such a structural organization may not be universal to vertebrate cells, but rather reflect an endothelial adaptation of the Golgi ribbon to the need for WPB production, providing a flexible system capable of modulating the size of these organelles in response to physiological and pharmacological cues [29, 34, 39, 41]. However, the abovementioned indirect evidence of similar dimers being present in HeLa cells [43] suggests that this might be a widespread feature. In the “breathing” ribbon model (Fig. 5C), formation of a dimer prevents a similar strong interaction of either of the constituent ministacks with another monomer. This implies that the machinery responsible for stable dimer formation must segregate at the location of the ministacks' interaction (Additional file 5: Fig. S5); that is, at a site on the rims of the cisternae. Polarization of molecular machinery within organelles has been described. For instance, MyRIP, which interacts with Factin, localizes at one tip of WPBs [36]. In ministack monomers, Golgi structural proteins localize at the rims of the cisternae [47] and one possibility is that lipidmediated phase separation and consequent protein spatial restriction [48] may segregate the machinery involved in dimer stabilization.
Our modelingbased prediction of functionally distinct and stable ministack dimers does not shed light on the physical nature of their embedding into the ribbon superstructure. Fluorescence recovery after photobleaching (FRAP) experiments firmly established that cisternal membranes of neighboring ministacks are continuous [16, 17, 49]. Therefore, a question raised by our “breathing” model of the Golgi ribbon is that of what molecular or physicochemical characteristics locally define the ministack dimers, while allowing their membrane continuity with adjacent ministacks. Our findings may prompt further studies to investigate this issue.
Limitations of the study
Being based on incompletely defined biological processes, the simplistic modeling exercises deployed in the present study may incompletely capture the biological complexity underlying the phenomena we formalized mathematically.
For example, the Golgi ribbon models here presented are deduced exclusively by the behavior of a single biological system, WPBs. This is due to the fact that, to our knowledge, this is the only system for which a relationship between the structure of the biosynthetic compartment (the Golgi ribbon) and size of secretory output (WPBs) is established [29, 32]. In addition, while our model of WPB biogenesis has withstood several experimental tests [29, 32, 39], it does not  for example  consider whether, given their size, limited lateral mobility of VWF quanta within the continuous lumen of the TGN may favor the observed high frequency of 2QWPBs. Actually, timelapse imaging indicates that VWF aggregations within the Golgi are very mobile (Movie S1 in reference [29]). However, due to limitations in resolution it is not clear whether this VWF mobility is inherent to quanta or rather reflects the movement of VWF quantumcontaining portions of the ribbon relative to each other. Future, more sophisticated imaging approaches may help decipher the nature of the VWF mobility observed at the Golgi and, if relevant, lead to a refined WPB biogenesis model. Finally, VWF residence time at the TGN, determined by the balance between its rates of arrival (anterograde traffic) at, and of departure from (WPB budding) the TGN, may impact quanta copackaging. VWF kinetics at the Golgi is however poorly defined and therefore not considered in our WPB biogenesis model.
Our models of the Golgi ribbon may also be oversimplified. In endothelial cells, the TGN is present as a continuous compartment spanning adjacent ministacks within the ribbon [29]. This structural organization of the TGN, embedded in our WPB biogenesis model, is also assumed with respect the Golgi structures we hypothesize in the present study. However, it is not established whether TGN continuity is a general feature of all vertebrate cell types. Also, the intermediate compartment (IC) and the recycling endosome (RE) networks were recently proposed to participate in Golgi ribbon formation by providing stable templates for the organization of ministacks into the ribbon arrangement [50]; this model of ribbon organization was not considered in our work.
Finally, our modeling of binding rates may not be accurate since quantitation of ministack monomers and dimers was possible only in nocodazole treated cells. Within an assembled ribbon, the presence of microtubules and/or other molecular components may impact binding rates in ways we have not accounted for here.
Conclusions
Our modeling exercise has provided insight into the dynamics of the smaller class of WPBs. Further detailed analysis  using the quantitative approach that we introduce in this study  of the roles of the known endothelial exocytic machinery would identify the molecular mechanisms regulating basal secretion and thus just how the 1QWPBs are preferentially released; a process potentially critical to setting the level of plasma VWF and thus thresholds for hemostasis, fundamental to both physiology and pathology [35].
Of more general interest and importance are the implications of this work regarding the Golgi apparatus and its architecture. Despite the mentioned limitations and the assumptions embedded in our mathematical models, these led us to the identification of stable, microtubuleindependent, ministack dimers as a major component of Golgi apparatus. The “breathing” ribbon model raises obvious questions as to the molecular machinery involved in stable ministack dimer formation and ministack dynamics in the context of the ribbon. If this model does indeed reflect an actual structural arrangement, a systematic approach designed to identify the machinery controlling the architecture of the Golgi ribbon, taking advantage of WPB size distribution as proxy of the ministack linkage status, could further our understanding of how this remarkable structure is formed and, potentially, reveal how its fragmentation is linked to disease.
Methods
Cells
Human umbilical vein endothelial cells (HUVECs) pooled from donors of both sexes were obtained commercially from PromoCell or Lonza. Cells were expanded and used within 15 population doublings, corresponding to passage 3 to 4 after expansion by our lab and maintained in HUVEC Growth Medium (HGM): M199 (Gibco, Life Technologies), 20% Fetal Bovine Serum, (Labtech), 30 μg/mL endothelial cell growth supplement from bovine neural tissue and 10 U/mL Heparin (both from SigmaAldrich). Cells were cultured at 37 °C, 5% CO_{2}, in humidified incubators.
Reagents
Chemicals
All common reagents were from Merck (SigmaAldrich). Nocodazole (M1404), Phorbol 12myristate 13acetate (P8139) and DMSO (D2650) were from Merck (SigmaAldrich). Nocodazole and PMA were dissolved in DMSO to generate 10 and 0.1 mg/mL stock solutions, respectively. Antibodies used in this study were rabbit polyclonal antibody raised against the cleaved VWF propeptide region [51]; mouse monoclonal antiVEcadherin (BD Bioscience, Clone 557H1, cat. no. 555661); sheep antiTGN46 (BioRad, cat. No, AHP500G). siRNAs targeting firefly Luciferase (5’CGUACGCGGAAUACUUCGAdTdT3’), human VWF (5’GGGCUCGAGUGUACCAAAAdTdT3’) and human MyRIP (5’AAGGTGGGAATTATTATTTAA3’ and 5’CCAAATTTACTTCCCAATAAA3’) were previously described and validated [36, 52, 53]. siRNAs were custom synthesized by Eurofins MWG Operon.
Treatments
HUVECs were seeded on gelatincoated NUNC 96well plates (Thermo Fisher, cat.no. 167008) at 1500020000 cells/well and maintained in HGM. Untreated cells were fixed after 48 h. In WPB repopulation experiments, 23 h after seeding, cells were fed with HGM supplemented with either DMSO or with PMA (100 ng/mL) for 1 h to elicit massive VWF secretion and deplete WPBs, followed by nocodazole incubation (1 μg/mL) for 24 h to allow WPB formation from separated Golgi ministacks. In the experiments involving VWF or MyRIP depletion, the indicated amounts of siRNAs were electroporated into one million HUVECs per reaction with an AMAXA Nucleofector II (Lonza), following the manufacturer’s instructions. Cells were then plated in 96well plates as described above and cultured for 48 h. Cells were then rinsed twice with warm, fresh HGM and fixed by incubation with 4% formaldehyde in PBS (10 min, at room temperature).
Immunostaining and image acquisition
After permeabilization (with 0.2% TX100 in PBS) and blocking (with 1 % bovine serum albumin in PBS), cells were incubated with primary antibodies, followed by secondary Alexa Fluor dyeconjugated antibodies (Life Technologies), both diluted in 1% BSA, 0.02% TX100 in PBS. Nuclei were counterstained with Hoechst 33342 (Life Technologies, H3570). Imaging was carried out with an Opera High Content Screening System (Perkin Elmer), using a 40x air objective (NA 0.6). Nine fields of view per well in 816 replicate wells were acquired, corresponding to approximately 24004800 cells per condition. The size in μm of ~ 100000 WPBs was measured for each condition. For Golgi and cytoplasm WPB size determination, together with WPBs the Golgi apparatus and cell boundaries were visualized with antiTGN46 and VEcadherin antibodies. Imaging was carried out with a 63x (NA 1.3) oil immersion objectives on a Leica Microsystems TCS SP5 confocal system. Eleven stacks (0.5 μm zstep) were reduced to maximum intensity projections. Approximately, 6000 to 9000 WPBs were analyzed for each of three independent experiments.
Image analysis
Highthroughput morphometry (HTM), the analytical pipeline including image processing, organelle segmentation and measurement of morphological parameters, has been described in detail elsewhere [29]. Briefly, images of WPB (visualized with antiVWF propeptide antibody) were segmented using Python (v2.7) (see script in Additional file 8). Data analysis was done using R (i386 3.1.0; script in Additional file 10). The size in μm of ~ 100000 WPBs was measured for each condition. WPB lengths were expressed as numbers of quanta as described in the Results section. For WPBs localized at the Golgi or in the cytoplasm, three experiments were carried out; WPB were segmented using a FIJI macro (Additional file 9) and their size analyzed with R, as described above.
Electron microscopy
Nocodazoletreated cells were highpressure frozen, freesubstituted (HPF/FS) and processed as described previously [54]. Samples were imaged with FEI Tecnai 20 TEM using an OlympusSoft Imaging System Morada camera.
Sample size
For each condition the minimum number of cells imaged was from 72 fields of view (FOV), corresponding to 8 wells of 96well plates. The number of independent biological replicates for each condition was 2 or 3, depending on the condition. To rule out experimenter’s bias, image acquisition (i.e., selection of FOV, etc.) and measurements taken from the images were fully automated. In pilot experiments (not reported in the present study), we sampled approximately 100000 WPB from control and treated cells and determined the statistical difference between these populations. We then took random samples of diminishing numbers of WPBs from these two groups and determined the point (10000 organelles) at which the significance of the difference fell below P ≤ 0.05. Using nocodazole to shorten the sizes of WPBs, and sampling as described above, the statistical power is 1 (the effect size of this treatment is approximately 2.02 measured by Cohen’s d). For the analyses reported in the present study, we therefore collected data from ~ 100000 WPBs for each condition.
Mathematical modeling
To create the model fit for Fig. 3A, we used Eq. 3, for n ≥ 2, (see text). We initially ignored the frequency of 1QWPBs. The relative frequencies of WPBs of lengths 2Q, 3Q, 4Q… are then given by 1p, p(1p), p^{2}(1p), … For multiple values of p in the range [0,1], we generated this distribution and measured the KolmogorovSmirnoff distance (this is the maximal absolute difference between the cumulative distribution functions) to the experimental relative frequency distribution of WPBs of lengths 2Q, 3Q, 4Q… We found the value of p that minimized this distance. Then, using Eq. 3, for n = 1, (see text) with the optimal value of p already found, we fixed α to give us the exactly correct frequency of 1QWPBs.
To create the model distributions in Figs. 5A and B, we fixed values for d_{1} and d_{2} and then ran a code (Additional file 7) to numerically simulate the stochastic process described by the reactions given in Additional file 6: A. At a late time (t = 5000), we output the relative frequencies of the Golgi portions of length 15 ministacks. For Fig. 5B, we then computed the KolmogorovSmirnoff distance to the experimental values of those relative frequencies. We repeated the process for d_{1} = 1, 2, 3, …., 20 and d_{2} = 1, 2, …., 400 and found the values of d_{1} and d_{2} that minimized the distance. Note that had these values been at an extreme value of the range, we would have searched a broader set of values for d_{1} and/or d_{2}.
To create the model fits for Fig. 4C, we fit values of p to the data from the control and three siRNA concentration experiments, as we did for Fig. 3A. We then used the frequencies of 1QWPBs to exactly determine α for each experiment. However, we assumed that α would not vary due to siRNA treatment, so that its values should be the same for all conditions. Thus, we took the weighted mean of the values of α that best fit each data set. We multiplied the value of α obtained for the luciferase siRNA condition by the total number of WPBs in that experiment and added the values of α obtained for VWF siRNA at 10200 pmol, multiplied by the corresponding number of WPBs in each condition, then divided by the total number of WPBs over all four conditions. We then plotted the model predicted distributions with that single value of α and the values of p chosen for each experiment separately. Note that had we allowed α to vary between the conditions, the fits would have been considerably better.
Availability of data and materials
The relevant data generated and analyzed in this study are included in this published article and its supplementary information. The original datasets used and analyzed for the present study are available from KMP and DFC on request.
Notes
The infinite ribbon assumption is made to avoid celldependent variations; to reflect, like the WPB distribution, a cell population; and also to simplify probability calculations.
We note that the best fit value of α decreased with increasing siRNA concentration; data not shown.
Abbreviations
 WPB:

WeibelPalade Body
 VWF:

Von Willebrand Factor
 TGN:

transGolgi Network
 ER:

endoplasmic reticulum
References
Vildanova MS, Wang W, Smirnova EA. Specific organization of Golgi apparatus in plant cells. Biochem (Mosc). 2014;79(9):894–906.
Yano H, YamamotoHino M, Abe M, Kuwahara R, Haraguchi S, Kusaka I, et al. Distinct functional units of the Golgi complex in Drosophila cells. Proc Natl Acad Sci USA. 2005;102(38):13467–72.
Sato M, Saegusa K, Sato K, Hara T, Harada A, Sato K. Caenorhabditis elegans SNAP29 is required for organellar integrity of the endomembrane system and general exocytosis in intestinal epithelial cells. Mol Biol Cell. 2011;22(14):2579–87.
Nakamura N, Wei JH, Seemann J. Modular organization of the mammalian Golgi apparatus. Curr Opin Cell Biol. 2012;24(4):467–74.
Wei JH, Seemann J. Unraveling the Golgi ribbon. Traffic (Copenhagen, Denmark). 2010;11(11):1391–400.
Wei JH, Seemann J. Golgi ribbon disassembly during mitosis, differentiation and disease progression. Curr Opin Cell Biol. 2017;47:43–51.
Thayer DA, Jan YN, Jan LY. Increased neuronal activity fragments the Golgi complex. Proc Natl Acad Sci USA. 2013;110(4):1482–7.
Ireland S, Ramnarayanan S, Fu M, Zhang X, Zhang J, Li J, et al. Cytosolic Ca(2+) Modulates Golgi Structure Through PKCalphaMediated GRASP55 Phosphorylation. iScience. 2020;23(3):100952.
Daussy CF, Monard SC, Guy C, MunozGonzalez S, Chazal M, Anthonsen MW, et al. The inflammasome components NLRP3 and ASC act in concert with IRGM to rearrange the Golgi during HCV infection. J. Virol. 2021;95(3):e0082620.
Gosavi P, Houghton FJ, McMillan PJ, Hanssen E, Gleeson PA. The Golgi ribbon in mammalian cells negatively regulates autophagy by modulating mTOR activity. J Cell Sci. 2018;131(3):jcs211987.
FarberKatz SE, Dippold HC, Buschman MD, Peterman MC, Xing M, Noakes CJ, et al. DNA damage triggers Golgi dispersal via DNAPK and GOLPH3. Cell. 2014;156(3):413–27.
Bellouze S, Baillat G, Buttigieg D, de la Grange P, Rabouille C, Haase G. Stathmin 1/2triggered microtubule loss mediates Golgi fragmentation in mutant SOD1 motor neurons. Mol Neurodegener. 2016;11(1):43.
Rabouille C, Haase G. Editorial: Golgi Pathology in Neurodegenerative Diseases. Front Neurosci. 2015;9:489.
Joshi G, Chi Y, Huang Z, Wang Y. Abetainduced Golgi fragmentation in Alzheimer's disease enhances Abeta production. Proc Natl Acad Sci USA. 2014;111(13):E1230–9.
Gosavi P, Gleeson PA. The Function of the Golgi Ribbon Structure  An Enduring Mystery Unfolds! Bioessays. 2017;39(11). https://doi.org/10.1002/bies.201700063.
Feinstein TN, Linstedt AD. GRASP55 regulates Golgi ribbon formation. Mol Biol Cell. 2008;19(7):2696–707.
Puthenveedu MA, Bachert C, Puri S, Lanni F, Linstedt AD. GM130 and GRASP65dependent lateral cisternal fusion allows uniform Golgienzyme distribution. Nat Cell Biol. 2006;8(3):238–48.
Yadav S, Puthenveedu MA, Linstedt AD. Golgin160 recruits the dynein motor to position the Golgi apparatus. Dev Cell. 2012;23(1):153–65.
Zhang Y, Seemann J. Rapid degradation of GRASP55 and GRASP65 reveals their immediate impact on the Golgi structure. J Cell Biol. 2021;220(1):e202007052.
Witkos TM, Lowe M. The Golgin Family of CoiledCoil Tethering Proteins. Frontiers in cell and developmental biology. 2015;3:86.
KulkarniGosavi P, Makhoul C, Gleeson PA. Form and function of the Golgi apparatus: scaffolds, cytoskeleton and signalling. FEBS Lett. 2019;593(17):2289–305.
Thyberg J, Moskalewski S. Microtubules and the organization of the Golgi complex. Exp Cell Res. 1985;159(1):1–16.
Miller PM, Folkmann AW, Maia AR, Efimova N, Efimov A, Kaverina I. Golgiderived CLASPdependent microtubules control Golgi organization and polarized trafficking in motile cells. Nat Cell Biol. 2009;11(9):1069–80.
Tang D, Zhang X, Huang S, Yuan H, Li J, Wang Y. MenaGRASP65 interaction couples actin polymerization to Golgi ribbon linking. Molecular biology of the cell. 2016;27(1):137–52.
Makhoul C, Gosavi P, Duffield R, Delbridge B, Williamson NA, Gleeson PA. Intersectin1 interacts with the golgin GCC88 to couple the actin network and Golgi architecture. Mol Biol Cell. 2019;30(3):370–86.
Bailey Blackburn J, Pokrovskaya I, Fisher P, Ungar D, Lupashin VV. COG Complex Complexities: Detailed Characterization of a Complete Set of HEK293T Cells Lacking Individual COG Subunits. Front Cell Dev Biol. 2016;4:23.
Gordon DE, Bond LM, Sahlender DA, Peden AA. A targeted siRNA screen to identify SNAREs required for constitutive secretion in mammalian cells. Traffic (Copenhagen, Denmark). 2010;11(9):1191–204.
Liu S, Storrie B. How Rab proteins determine Golgi structure. Int Rev Cell Mol Biol. 2015;315:1–22.
Ferraro F, KristonVizi J, Metcalf DJ, MartinMartin B, Freeman J, Burden JJ, et al. A twotier Golgibased control of organelle size underpins the functional plasticity of endothelial cells. Dev Cell. 2014;29(3):292–304.
McCormack JJ, Lopes da Silva M, Ferraro F, Patella F, Cutler DF. WeibelPalade bodies at a glance. J Cell Sci. 2017;130(21):3611–7.
LuiRoberts WW, Collinson LM, Hewlett LJ, Michaux G, Cutler DF. An AP1/clathrin coat plays a novel and essential role in forming the WeibelPalade bodies of endothelial cells. J Cell Biol. 2005;170(4):627–36.
Ferraro F, Lopes da Silva M, Grimes W, Lee HK, Ketteler R, KristonVizi J, et al. WeibelPalade body size modulates the adhesive activity of its von Willebrand Factor cargo in cultured endothelial cells. Sci Rep. 2016;6:32473.
Yuan L, Chan GC, Beeler D, Janes L, Spokes KC, Dharaneeswaran H, et al. A role of stochastic phenotype switching in generating mosaic endothelial cell heterogeneity. Nat Commun. 2016;7:10160.
McCormack JJ, HarrisonLavoie KJ, Cutler DF. Human endothelial cells sizeselect their secretory granules for exocytosis to modulate their functional output. J Thromb Haemost. 2020;18(1):243–54.
Lopes da Silva M, Cutler DF. von Willebrand factor multimerization and the polarity of secretory pathways in endothelial cells. Blood. 2016;128(2):277–85.
Nightingale TD, Pattni K, Hume AN, Seabra MC, Cutler DF. Rab27a and MyRIP regulate the amount and multimeric state of VWF released from endothelial cells. Blood. 2009;113(20):5010–8.
Cole NB, Sciaky N, Marotta A, Song J, LippincottSchwartz J. Golgi dispersal during microtubule disruption: regeneration of Golgi stacks at peripheral endoplasmic reticulum exit sites. Mol Biol Cell. 1996;7(4):631–50.
Starke RD, Paschalaki KE, Dyer CE, HarrisonLavoie KJ, Cutler JA, McKinnon TA, et al. Cellular and molecular basis of von Willebrand disease: studies on blood outgrowth endothelial cells. Blood. 2013;121(14):2773–84.
Ferraro F, Patella F, Costa JR, Ketteler R, KristonVizi J, Cutler DF. Modulation of endothelial organelle size as an antithrombotic strategy. J Thromb Haemost. 2020.
Ketteler R, Freeman J, Ferraro F, Bata N, Cutler DF, KristonVizi J, et al. Imagebased siRNA screen to identify kinases regulating WeibelPalade body size control using electroporation. Sci Data. 2017;4:170022.
LopesdaSilva M, McCormack JJ, Burden JJ, HarrisonLavoie KJ, Ferraro F, Cutler DF. A GBF1Dependent Mechanism for Environmentally Responsive Regulation of ERGolgi Transport. Dev Cell. 2019;49(5):786–801 e786.
Stevenson NL, MartinMartin B, Freeman J, KristonVizi J, Ketteler R, Cutler DF. G proteincoupled receptor kinase 2 moderates recruitment of THP1 cells to the endothelium by limiting histamineinvoked WeibelPalade body exocytosis. J Thromb Haemost. 2014;12(2):261–72.
Kondylis V, van Nispen tot Pannerden HE, Herpers B, FriggiGrelin F, Rabouille C. The golgi comprises a paired stack that is separated at G2 by modulation of the actin cytoskeleton through Abi and Scar/WAVE. Dev Cell. 2007;12(6):901–15.
Ravichandran Y, Goud B, Manneville JB. The Golgi apparatus and cell polarity: Roles of the cytoskeleton, the Golgi matrix, and Golgi membranes. Curr Opin Cell Biol. 2020;62:104–13.
Yadav S, Puri S, Linstedt AD. A primary role for Golgi positioning in directed secretion, cell polarity, and wound healing. Mol Biol Cell. 2009;20(6):1728–36.
Bisel B, Wang Y, Wei JH, Xiang Y, Tang D, MironMendoza M, et al. ERK regulates Golgi and centrosome orientation towards the leading edge through GRASP65. J Cell Biology. 2008;182(5):837–43.
Tie HC, Ludwig A, Sandin S, Lu L. The spatial separation of processing and transport functions to the interior and periphery of the Golgi stack. Elife. 2018;7.
King C, Sengupta P, Seo AY, LippincottSchwartz J. ER membranes exhibit phase behavior at sites of organelle contact. Proc Natl Acad Sci USA. 2020;117(13):7225–35.
Cole NBS, Smith CL, Sciaky N, Terasaki M, Edidin M, LippincottSchwartz J. Diffusional mobility of Golgi proteins in membranes of living cells. Sci (New York, NY). 1996;273(5276):797–801.
Saraste J, Prydz K. A New Look at the Functional Organization of the Golgi Ribbon. Front Cell Dev Biol. 2019;7:171.
Hewlett L, Zupancic G, Mashanov G, Knipe L, Ogden D, Hannah MJ, et al. Temperaturedependence of WeibelPalade body exocytosis and cell surface dispersal of von Willebrand factor and its propolypeptide. PLoS One. 2011;6(11):e27314.
Elbashir SM, Harborth J, Lendeckel W, Yalcin A, Weber K, Tuschl T. Duplexes of 21nucleotide RNAs mediate RNA interference in cultured mammalian cells. Nature. 2001;411(6836):494–8.
Starke RD, Ferraro F, Paschalaki KE, Dryden NH, McKinnon TA, Sutton RE, et al. Endothelial von Willebrand factor regulates angiogenesis. Blood. 2011;117(3):1071–80.
Zenner HL, Collinson LM, Michaux G, Cutler DF. Highpressure freezing provides insights into WeibelPalade body biogenesis. J Cell Sci. 2007;120(Pt 12):2117–25.
Acknowledgements
N/A
Funding
This work was funded by the Medical Research Council (grants MC_UU_12018/2 and MC_UU_00012/2 to D.F.C.) and by the British Heart Foundation (grant PG/14/76/31087 to D.F.C.).
Author information
Authors and Affiliations
Contributions
Conception: KMP, FF, DFC. Design: KMP, JJM, MLP, FP, FF, DFC. Acquisition of data: JJM, FP, JJB, KHL, YYBQ, DS. Analysis of data: KMP, JJM, MLS, FP, JJB, FF, DFC. Interpretation of data: KMP, JJM, MLS, FP, JJM, FF, DFC. Drafted or revised the work: KMP, JJM, MLP, JJB, FF, DFC. All authors have approved this submission and acknowledge their responsibilities arising from this.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Supplementary Information
Additional file 1: Fig. S1.
WPB size distribution measured from a population of endothelial cells. A WPB size distribution as measured by highthroughput morphometry (see Methods). B HUVECs show variability in VWF expression and WPB production. WPBs and nuclei were visualized as indicated in Fig. 1B; the Golgi was visualized with an antiGM130 antibody; scale bar: 20 μm.
Additional file 2: Fig. S2.
Further simulations based on the “miniribbon collection” and the “free ministack dimers plus ribbon” models. A Predictions of WPB size distributions at different probabilities of quantum ministack occupancy probability (p), where the Golgi is made exclusively of ribbons of length 4l. B A Golgi made of ribbons and ministack dimers, where the latter are present in proportion q; different q values are depicted. C Based on the model in B, expected WPB size distributions (stars) were calculated for the indicated ministack occupancy probability (p) and the proportion of the Golgi formed by dimers (q) and compared to the measured WPB size distribution (blue bars).
Additional file 3: Fig. S3.
Shifts in WPB size distribution following increase in basal secretion. A Efficiency of MyRIP knockdown (300 pmol of each siRNA were used per reaction); means and ranges from two independent experiments are reported; Student’s ttest. B Morphometric analysis of WPB size shows that MyRIP knockdown increases the fraction of long organelles (defined as those > 2 μm); data shown are from one of the replicate experiments shown in A; statistical analysis was nonparametric (MannWhitney test).
Additional file 4: Fig. S4.
Shifts in WPB size distribution following Golgi ribbon unlinking. A HUVECs were treated for 1 h with either DMSO or 100 ng/mL PMA and processed for immunofluorescence. PMA almost completely depletes the cellular pool of WPBs. Scale bar, 25 μm. B Representative micrograph of HUVECs pretreated with PMA as in A to clear WPBs and then chased in nocodazole for 24 h; scale bar: 25 μm. C Size distribution of WPBs 24 h after nocodazole treatment, following pretreatment with either DMSO (control) or PMA to deplete the organelles. The resulting populations of organelles following 24 h nocodazole treatment are newlymade WPBs plus those left after basal exocytosis, in the case DMSO; almost completely newlymade WPBs, in the case of PMA (see panel A). D Visualization of a hypothetical arrangement of the Golgi ribbon where stable ministacks dimers are independent of microtubules.
Additional file 5: Fig. S5.
Modeling of ministack linking to form the Golgi ribbon. The possible ministacks associations up to pentamers are depicted (see text).
Additional file 6.
A Reaction scheme for ministack linking and unlinking. B Equation system 4.
Additional file 7.
Matlab code for numerical simulations of the reaction scheme described in Additional file 5 A.
Additional file 8.
Python script for WPB segmentation.
Additional file 9.
FIJI macro for Golgi and extraGolgi WPB segmentation.
Additional file 10:
R script for WPB size determination from images.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.
About this article
Cite this article
Page, K.M., McCormack, J.J., LopesdaSilva, M. et al. Structure modeling hints at a granular organization of the Golgi ribbon. BMC Biol 20, 111 (2022). https://doi.org/10.1186/s12915022013053
Received:
Accepted:
Published:
DOI: https://doi.org/10.1186/s12915022013053