TANGO 1 regulates membrane tension to mediate procollagen 1 export 2 3

ABSTRACT The endoplasmic reticulum (ER)-resident transmembrane protein TANGO1 assembles into rings around COPII subunits at ER exit sites (ERES), and links cytosolic membrane-remodeling machinery, tethers, and ER-Golgi intermediate compartment (ERGIC) membranes to procollagens in the ER lumen (Raote et al., 2018). This arrangement is proposed to create a direct route for transfer of procollagens from ERES to ERGIC membranes. Here, we present a physical model in which TANGO1 forms a linear filament that wraps around COPII lattices at ERES to stabilize the neck of a growing carrier on the cytoplasmic face of the ER. Importantly, our results show that TANGO1 can induce the formation of transport intermediates by regulating ER membrane tension. Altogether, our theoretical approach provides a mechanical framework of how TANGO1 acts as a membrane tension regulator to control procollagen export from the ER.


25
The endoplasmic reticulum (ER)-resident transmembrane protein TANGO1 assembles 26 into rings around COPII subunits at ER exit sites (ERES), and links cytosolic membrane-27 remodeling machinery, tethers, and ER-Golgi intermediate compartment (ERGIC) mem-28 branes to procollagens in the ER lumen (Raote et al., 2018). This arrangement is proposed 29 to create a direct route for transfer of procollagens from ERES to ERGIC membranes. 30 Here, we present a physical model in which TANGO1 forms a linear filament that wraps 31 around COPII lattices at ERES to stabilize the neck of a growing carrier on the cytoplas-32 mic face of the ER. Importantly, our results show that TANGO1 can induce the formation 33 of transport intermediates by regulating ER membrane tension. Altogether, our theoret-34 ical approach provides a mechanical framework of how TANGO1 acts as a membrane 35 tension regulator to control procollagen export from the ER. 36 Multicellularity requires not only the secretion of signaling proteins -such as neurotransmitters, 38 cytokines, and hormones-to regulate cell-to-cell communication, but also of structural proteins 39 such as collagens, which form basement membranes and more generally the extracellular ma-40 trix (ECM) (Kadler et al., 2007;Mouw, Ou and Weaver, 2014). These extracellular assemblies 41 of collagens are necessary for skin biogenesis and to form the connective tissues. ECM also 42 likely acts as a ruler to control the size of a tissue. Collagens, like all secretory proteins, contain 43 a signal sequence that targets their de novo synthesis into the endoplasmic reticulum (ER). After 44 their glycosylation, folding and trimerization, the bulky procollagens are exported from the ER 45 to the Golgi complex and thence to the exterior of the cells. The export domains of secretory 46 cargoes, named the ER exit sites (ERES), are a fascinating subdomain of the ER, but the basic 47 understanding of how these domains are created and segregated from rest of the ER for the 48 purpose of cargo export still remains a major challenge. The discovery of TANGO1 as a key 49 player that sits at ERES has made the process of procollagen export and the organization of 50 ERES amenable to molecular analysis (Bard et al., 2006;Saito et al., 2009;Wilson et al., 2011). 51 In the lumen of the ER, The SH3 domain of TANGO1 binds procollagen via HSP47 (Saito et  high lateral spatial resolution using stimulated emission depletion (STED) nanoscopy in mam-62 malian tissue cultured cells (Raote et al., 2017(Raote et al., , 2018. These studies revealed that TANGO1 63 organizes at the ERES into ring-like structures, of ~200 nm in diameter, that corral COPII com-64 ponents. Moreover, an independent study showed that TANGO1 rings are also present in Dro-65 sophila melanogaster (Liu et al., 2017). 66 To further extend these findings, we combined STED nanoscopy with genetic manipulations 67 and established that TANGO1 rings are organized by (i) lateral self-interactions amongst 68 TANGO1-like proteins, (ii) radial interactions with COPII subunits, and (iii) tethering of small 69 ER-Golgi intermediate compartment (ERGIC) vesicles to assist in the formation a procollagen-70 containing transport intermediate (Raote et al., 2018). Overall, the accumulated data suggest a 71 mechanism whereby TANGO1 assembles into a functional ring, which selectively gathers and 72 organizes procollagen, remodels the COPII budding machinery, and recruits ERGIC mem-73 branes for the formation of a procollagen-containing transport intermediate. However, the bio-74 physical mechanisms governing these events and how they are regulated by TANGO1 remain 75 unknown. 76 Here, we present and analyze a biophysical model of TANGO1 ring assembly around polymer-77 izing COPII-coated structures. Our model allows us to address: (i) the physical mechanisms by 78 which TANGO1 and its interactors assemble into functional rings at ERES, forming a fence 79 around COPII coat components; and (ii) how TANGO1 fence can couple membrane tension in 80 two compartments to modulate the formation of carriers at the ERES. Overall, we propose a 81 novel mechanism of TANGO1-regulated procollagen export, which consists of two sequential 82 steps. First, TANGO1 rings, at the edge of a polymerizing COPII structure, stabilize the neck 83 of a growing procollagen-containing transport export intermediate and thus prevent premature 84 carrier fission. Second, carrier growth can be stimulated by the ability of TANGO1 to act as a 85 membrane tension regulator by tethering ERGIC membranes. Importantly, we show that 86 TANGO1-mediated local reduction of the membrane tension at the ERES reduces the energy 87 barrier required for carrier growth.  Experimental basis and assumptions of the model 107 To assess and rationalize the mechanisms by which TANGO1 assembles into rings at ERES, 108 we propose a physical model built on accumulated experimental data. 109 110 First, we hypothesize that TANGO1 forms a filament that can be held together by lateral pro-111 tein-protein interactions between TANGO1-family proteins (TANGO1, cTAGE5 and 112 TANGO1-Short) (Raote et al., 2018). This hypothesis is based on the following observations: or the binding affinity of TANGO1 for the COPII subunits is relatively large, the filament will 172 easily adapt its shape by wrapping around COPII patches forming a TANGO1 ring (a process  173 we refer to as ERES wetting) ( Figure 1B). As a result, there will be a linactant effect of 174 TANGO1 on COPII-coated ERES that will reduce the line energy, thus limiting the growth of 175 the ERES and the size of the TANGO1 rings ( Figure 1B). By contrast, if TANGO1 filaments 176 are very rigid or the affinity of TANGO1 proteins for COPII subunits is low (for instance, in 177 cells expressing mutants of TANGO1 with reduced or abrogated interaction to COPII proteins), 178 ERES wetting by the filament will be energetically unfavorable and as a results TANGO1 will 179 not act as a COPII linactant ( Figure 1B). 180 181 To quantitatively analyze this hypothesis, we start by considering a two-dimensional scenario 182 where both TANGO1 filaments and COPII coats lie on the plane of a flat two-dimensional 183 membrane (the role of the membrane curvature and the three-dimensional organization of the 184 different molecular players to form a transport intermediate is described in the second part of 185 this article). We use a coarse-grained, continuum model, which implicitly considers TANGO1 186 family proteins (TANGO1, cTAGE5 and TANGO1-Short) and TANGO1-binding COPII sub-187 units. Here, the "microscopic" interaction energies are averaged out into "macroscopic" free 188 energies, such as the filament bending energy, or the coat line energy. Although simplistic in 189 nature, this continuum model is a suitable choice for a semi-quantitative description of the main 190 physical mechanisms driving ring formation, as structural data on TANGO1 proteins are cur-191 rently lacking. 192 193 For the sake of simplicity, we consider in our physical model that the ER membrane contains a 194 certain number of independent, non-interacting COPII-enriched domains of radius , distrib-195 uted following a hexagonal array, with a center-to-center distance, , between domains ( Figure  196 S1A). To understand the effect of proteins of the TANGO1 family on the size and shape of 197 COPII domains along the ER membrane, we need to consider the different protein interactions 198 outlined above, namely (i) TANGO1-TANGO1 interactions, which control the bending energy 199 of the TANGO1 filament; (ii) TANGO1 interaction with peripheral COPII subunits, which 200 controls the line energy of the COPII domain; and (iii) TANGO1 interaction with regulatory 201 COPII proteins, which controls COPII polymerization kinetics. In sum, the total free energy of 202 the system is the addition of these different free energy terms (see Equations (M1-M4) in  terials and Methods, where a detailed mathematical description of the elastic model of 204 TANGO1 ring formation is presented). We consider that the total surface area of our system 205 and the total surface area covered by ERES is fixed, so instead of working with the extensive 206 free energy of the system, , we will work with the intensive free energy per unit ERES 207 area, = / )*)+ . This free energy density for a system of circular domains of radius can where is the wetting fraction, which represents the fraction of ERES boundary length asso-213 ciated with TANGO1 molecules; and is the radius of the TANGO1 ring. The first term of 214 Equation ( average ERES size corresponds to the minimum of the total free energy of the system Equation 259 (2) and determined the conditions promoting or preventing filament wrapping around COPII 260 patches, which we refer to as ERES wetting. This configuration of minimal free energy is ac-261 quired by optimizing the free parameters of the model, namely the dimensionless size of the 262 ERES, , and the wetting fraction, . 263 264 Since the free energy in Equation (2) has a linear dependence on the wetting fraction, , it is 265 monotonic with respect to this variable and therefore energy minimization will drive the system 266 to either complete wetting ( = 1), or complete dewetting ( = 0), depending on the sign of 267  (Figure 2A)  Computation of the preferred size of TANGO1 rings 304 TANGO1 rings surround COPII components (Raote et al., 2017), corresponding to a filament 305 full wetting condition (that is, = 1), as presented in Figure 1B (analysis of the ERES size in 306 dewetting conditions is presented in the Supplementary Information). Under wetting condi-307 tions, a ring of radius Z[\j is formed by a TANGO1 filament wrapping around a COPII patch. 308 The value of the optimal dimensionless ring size, Z[\j = Z[\j / 3 , is obtained by minimizing 309 Equation (2) in wetting conditions, which is equivalent to solve the fifth order algebraic equa-310 tion, 311 312 Because Equation (4) cannot be analytically solved, we opted to solve it numerically for dif-315 ferent values of the model's parameters. As a starting point, we took the parameter values ? = 316 500 ⋅ 5 (corresponding to the TANGO1 filaments having a persistence length of o ≃ 317 120 ), 3 = 100 , 3 = 0.05 (see Table 1), which yields ? CCC = 1. We then looked 318 for the solutions of Equation (4) as a function of the dimensionless coupling parameter, 3 F , and 319 of the relative line tension reduction, Δ CCCC . These results ( Figure 2B), show that ring formation 320 (wetting by the TANGO1 filament) can be induced by decreasing the coupling factor, 3 F , or by 321 increasing the linactant strength of TANGO1, Δ CCCC . Since Δ CCCC essentially corresponds to the 322 COPII-TANGO1 binding affinity, and hence our results indicate that TANGO1 rings are sta-323 bilized by the association of TANGO1 proteins with peripheral COPII subunits. Furthermore, 324 our results also show that the size of the TANGO1 rings decreases with increasing values of 325 Δ CCCC , and with increasing values of 3 F ( Figure 2B). Next, we computed the wetting-dewetting 326 diagram and the optimal TANGO1 ring size in wetting conditions as a function of the relative 327 line tension reduction, Δ CCCC , and of the bending rigidity, ? CCC ( Figure 2C, Figure  values of the filament spontaneous curvature, 3 F ( Figure 2B-D). In other words, both a large 335 affinity of TANGO1 proteins for COPII subunits and a small resistance of the TANGO1 fila-336 ment to bending (which in structural terms can be thought of as a small lateral protein-protein 337 interaction between the filament components) induce the formation of TANGO1 rings and tend 338 to reduce the size of these rings. 339 340 Comparison with experimental results 341 We previously reported that cells expressing mutants of TANGO1 with abrogated binding to 342 the COPII component Sec23 (TANGO1-∆PRD mutant) present both smaller and less stable 343 rings as compared to wild-type cells, including also the presence of some fused structures 344 (Raote et al., 2018). In cells expressing TANGO1-∆PRD, the interaction between one of the 345 filament components, TANGO1, and the COPII subunits is abolished, indicating that, although 346 a TANGO1 filament could still be formed -this mutant does not alter the interaction between 347 TANGO1 and other TANGO1 or cTAGE5 proteins (Raote et al., 2018)-, the filament should 348 be less line-active because the affinity to bind to the peripheral COPII subunits is reduced. In 349 this situation the filament proteins cTAGE5 (Saito et al., 2011(Saito et al., , 2014) and TANGO1-Short 350 (Maeda, Saito and Katada, 2016) can still bind Sec23 and therefore reduce, albeit to a lesser 351 extent than in wild-type cells, the COPII patch line energy. However, in our results presented 352 in Figure 2B-D and Figure S2, we observed that a reduction of the linactant strength of 353 TANGO1 (parameter Δ CCCC ) normally leads to an increase rather than a decrease of the ring size 354 (see supplementary information for a more detailed discussion). To investigate how the lack of 355 the PRD domain of TANGO1 contributes to form smaller rings, we explored how other differ-356 ential properties of TANGO1-∆PRD in relation to those of TANGO1-WT could lead to the 357 experimentally-observed reduction in ring sizes from about 275±70 nm to 170±65 nm (mean 358 Feret's diameter of the ring) (Raote et al., 2018). Our model predicts that the experimentally 359 observed reduction of TANGO1-∆PRD ring size needs to parallel either (i) spatio-temporal 360 regulation by the PRD of ERES dynamics (such as an increase in the parameter 3 F ); (ii) a re-361 duction of the filament bending rigidity, ? CCC; or (iii) an increase of the preferred curvature of 362 the filament, 3 F ( Figure 2B-D Can TANGO1 modulate the shape of a growing bud to accommodate large and complex car-374 goes? And, if so, would the TANGO1 ring structure be especially suited to achieve this task? 375 To answer these questions, we put together a physical model of transport intermediate for-376 mation that incorporates the effects of TANGO1 ring formation and wetting as discussed above. 377 In our model, we consider different scenarios under which TANGO1 can modulate the standard 378 spherical COPII carrier formation. 379 380 Qualitative description of TANGO1-mediated transport intermediate formation 381 The formation of the canonical coated transport carriers (such as COPI-, COPII-, or clathrin-382 coated carriers) relies on the polymerization of a large-scale protein structure on the membrane 383 surface, the protein coat. Polymerized coats usually adopt spherical shapes, which bend the 384 membrane underneath accordingly (Faini et al., 2013). Membrane bending is promoted if the 385 binding energy of the coat to the membrane is larger than the energy required to bend the mem-386 brane and if the coat structure is more rigid than the membrane (Kozlov et al., 2014;Saleem et 387 al., 2015). Hence, in the absence of a functional TANGO1, COPII coats generate standard 60-388 90 nm spherical transport carriers ( Figure 3A). In this situation, the neck of the growing carrier 389 prematurely closes without being able to fully incorporate long semi-rigid procollagen mole-390 cules, which are not efficiently recruited to the COPII export sites due to the lack of TANGO1 391 ( Figure 3A). In our model for TANGO1 ring formation, we proposed that one of the potential 392 roles of such a ring is to act as a linactant to stabilize free COPII subunits at the edge of the 393 polymerized structure (Glick, 2017;Raote et al., 2017) and hence prevent or kinetically delay 394 the premature closure of the bud neck ( Figure 3B). Moreover, mechanical forces pointing to-395 wards the cytosolic side of the bud, either from the ER lumen (e.g. TANGO1 pushing procol-396 lagen upwards) or from the cytosol (e.g. molecular motors pulling on the growing bud), will 397 induce the growth of the transport intermediate ( Figure 3D). Fusion of such vesicles to the budding 404 site would deliver membrane lipids to the ER membrane, which rapidly and transiently induces 405 a local drop in membrane tension, hence overcoming the tension-induced arrest in transport 406 intermediate growth ( Figure 3E). The shape and coat coverage of procollagen-containing ex-407 port intermediates remain, to the best of our knowledge, a matter of speculation. Both long 408 pearled tubes (Figure 3E)

437 438
Physical model of TANGO1-dependent transport intermediate formation 439 To quantitate the feasibility of the proposed pathway of transport intermediate growth ( Figure  440 3), we developed a physical model that accounts for the relative contribution of each of these 441 forces to the overall free energy of the system. Such a model allows us to predict the shape 442 transitions from planar membrane to incomplete buds and to large transport intermediates. In-443 tuitively, one can see that COPII polymerization favors the formation of spherical buds, 444 whereas TANGO1 linactant strength and filament bending prevent neck closure. Large out-445 ward-directed forces promote the growth of long intermediates, whereas large membrane ten-446 sions inhibit such a growth. Taking  we expand on this model to include the aforementioned contributions of TANGO1-like proteins 449 in modulating COPII-dependent carrier formation. We consider that the ER membrane is under 450 a certain lateral tension, 3 , and resists bending by a bending rigidity, r . Growth of a COPII 451 bud starts by COPII polymerization into a spherical shape of radius . The chemical potential 452 of the COPII coat, U , includes the COPII binding energy, U 3 , and the bending energy of the 453 underlying membrane (see Materials and Methods). As explained in the ring-formation model, 454 incomplete buds are associated with a line tension of the free subunits, 3 , which can be par-455 tially relaxed by the wetting of a TANGO1 ring, hence reducing the line tension by an amount 456 ∆ . In addition, we also consider the chemical potential of the TANGO1 ring, ? , which ac-457 counts for the filament assembly energy via lateral interactions, ? 3 , and the filament bending 458 energy. Next, we also account for the mechanical work of an outward-directed force, , which 459 favors transport intermediate growth. Finally, the fusion of incoming ERGIC53-containing 460 membranes is accounted by a sharp and local reduction in the lateral membrane tension, by an 461 amount equal to Δ . Altogether, we can write the total free energy per unit surface area with 462 respect to a naked flat membrane, U , as 463 464 where ‡ is the membrane surface area, U is the surface area of the membrane covered by the 467 COPII coat, o is the surface area of the carrier projection onto the flat membrane, is the 468 radius of the base of the carrier, ℎ is the height of the carrier, and ? is the radius of the 469 TANGO1 ring (Figure S3A, and Materials and methods section). We consider that the carrier 470 adopts the equilibrium configuration, corresponding to the shape of minimum free energy, 471 Equation (5). Although the system is not in equilibrium, this assumption will be valid as long 472 as the mechanical equilibration of the membrane shape is faster than the fluxes of the lipids and 473 proteins involved in the problem (  pression for the effective chemical potential, ", we can see that the application of a force in the 489 bud growth direction, , plays the same role as the coat binding free energy, U 3 , and therefore 490 helps counterbalance the elastic resistance of the membrane to deformation. In addition, the 491 lateral binding free energy of the TANGO1 filament, ? 3 , also helps, in wetting conditions, to 492 decrease the value of the effective coat line tension, " , thus preventing premature closure of the 493 bud neck (Figure 3).

495
Functional TANGO1 rings can control transport intermediate formation by force 496 exertion and membrane tension regulation 497 The free energy per unit area of the transport intermediate, U , has a non-trivial dependence on 498 the shape of the carrier, parametrized by the shape parameter, , as given by Equation (6). This 499 implies that multiple locally stable shapes, corresponding to different local minima of the free 500 energy, can coexist. To illustrate this dependence, the profile of the free energy per unit area, U , 501 as a function of the shape parameter , is shown for two different scenarios in Figure 4A. In 502 the first one, which corresponds to a situation where the COPII binding energy is relatively 503 small, U 3 = 0.012 ™ / 5 (top panel, Figure 4A), the global minimum of the free energy 504 corresponds to a shallow bud. Other locally stable shapes, corresponding to a shallow bud con-505 nected to a set of spheres, can be found. By contrast, in the second scenario illustrated in Figure  506 4A (bottom panel), which corresponds to a situation of relatively large COPII binding energy, 507 U 3 = 0.048 ™ / 5 , the transport intermediate will grow from an initially unstable shallow 508 bud (depicted in red, in Figure 4A, bottom panel) to a locally stable almost fully formed spher-509 ical carrier (depicted in yellow, in Figure 4A, bottom panel). Then, overcoming an energy bar-510 rier will result in further growth of the carrier into a large transport intermediate (depicted in  511 green, in Figure 4A, bottom panel). Next, we computed the profile of the free energy per unit 512 area, U , as a function of the shape parameter , for different values of the COPII binding en-513 ergy, U 3 , and of the TANGO1 bending rigidity, ? (Figure 4B). These results show that the 514 bending rigidity of the TANGO1 filament, when assembled around the growing COPII bud, 515 leads to the existence of a high energy barrier in the transition from a single bud to a multiple 516 bud transport intermediate, or pearled tube (Figure 4B, compare dashed lines corresponding to 517 a TANGO1 filament with no bending rigidity to the solid lines, where the TANGO1 filament 518 is associated with a certain bending rigidity and therefore resists bending). A transition could 519 still occur in this latter case, since the shape transition could occur through transient dewetting 520 of the TANGO1 filament or through intermediate shapes between a cylindrical tube and a set 521

538 539
We next looked for the locally and globally stable shapes of the transport intermediate, by com-540 puting the local minima of the overall energy of the system per unit area, Equation (6), for both 541 single buds (shape parameter < 1) or for long transport intermediates (shape parameter > 542 1). In Figure 5, we show, for a wide range of the model's parameters, the optimal shape of the 543 intermediate, as measured by the optimal shape parameter, * , and the corresponding free en-544 ergy per unit area for both single incomplete buds ( = 0 ; * < 1) (light blue lines in Figure  545 5) and long intermediates containing one full bud plus an incomplete bud ( = 1 ; 1 < * < 2) 546 (orange lines in Figure 5). Our results indicate that the rigidity of the TANGO1 filament has 547 no effect on the shape of the transport intermediate and does not trigger the elongation of the 548 COPII bud (Figure 5A). When we varied the effective coat line tension, fII (Figure 5B), we 549 observed that for large values of the effective line tension, the shape of the intermediate tends 550 to the complete bud ( = 1), but a transition to long pearled shapes is not promoted. In strong 551 contrast, the COPII coat binding energy, U 3 , does play an important role in controlling the 552 elongation of the carriers, since our results ( Figure 5C) show that increasing this value leads to 553 a sharp transition from shallow buds ( Figure 5C, top panel, solid blue line) to shallow pearled 554 tubes ( Figure 5C, top panel, solid orange line). Similarly, the application of a force directed 555 towards the cytosol at the tip of the growing intermediate also leads to the transition from a 556 shallow bud to a pearled tube (Figure 5D). 557 558 559 560  wards the cytosol (Figure 5D and Figure 6); and (iii) local reduction of the membrane tension 615 (Figure 6). TANGO1 can directly or indirectly control each of these possibilities (Ma and  616 Goldberg, 2016; Raote et al., 2018). Interestingly, the TANGO1 ring properties, such as the 617 linactant power of TANGO1 or the TANGO1 filament bending rigidity, are not drivers of the 618 incomplete bud to long transport intermediate transition (Figure 6C,D), but they seem to act 619 more as kinetic controllers of the transition by preventing bud closure (Figure 4). 620 621 622 623   issue on the morphology of the transport intermediates that shuttle procollagens form the ER to 693 the Golgi complex. 694 695 TANGO1 as a regulator of membrane tension homeostasis 696 We previously showed that TANGO1 forms circular ring-like structures at ERES surrounding 697 COPII components (Raote et al., 2017). We also revealed the interactions that are required for 698 TANGO1 ring formation, which are also important to control TANGO1-mediated procollagen 699 export from the ER (Raote et al., 2018). However, it still remained unclear whether and how 700 TANGO1 rings could organize and coordinate the budding machinery for efficient procollagen-701 export. Here, we proposed, described, and analyzed a feasible biophysical mechanism of how 702 TANGO1 mediates the formation of procollagen-containing transport intermediates at the ER. 703 The general idea backed by the results of our model is that TANGO1 rings serve as stabilizers 704 of small buds, preventing the premature formation of standard COPII coats. TANGO1 is ubiq-705 uitously expressed in mammalian cells, including cells that secrete very low amounts of colla-706 gen. Furthermore, TANGO1 resides in most ERES in all these different cell lines, yet small 707 COPII-coated vesicles form normally in those sites. How can this be understood? We propose 708 that the ability of TANGO1 to form rings around COPII subunits is a first requirement for 709 TANGO1 to promote procollagen export in non-standard COPII vesicles. Accumulations of 710 export-competent procollagen at the ERES could re-organize the TANGO1 molecules laying 711 there into functional rings surrounding COPII components and kinetically preventing the for-712 mation of small COPII carriers. Tethering of ERGIC53-containing vesicles mediated by the 713 TANGO1 TEER domain (Raote et al., 2018) could be the trigger to allow for carrier growth. 714 Importantly, the ER-specific SNARE protein Syntaxin18 and the SNARE regulator SLY1, 715 which together trigger membrane fusion at the ER, are also required for procollagen export in 716 a TANGO1-dependent manner (Nogueira et al., 2014). Fusion of ERGIC membranes to the 717 sites of procollagen export would lead to a local and transient reduction of the membrane ten-718 sion, which can promote, according to our theoretical results, the growth of the COPII carrier. 719 In this scenario, TANGO1 would act as a regulator of membrane tension homeostasis to control 720 procollagen export at the ERES. In parallel, we can also foresee a situation by which TANGO1 721 rings help pushing procollagen molecules into the growing carrier and couple this pushing force 722 to procollagen folding, through the chaperone HSP47 (Figure 3). This pushing force, according 723 to our model, would also promote the formation of a large intermediate and hence TANGO1  724 could act as a sensor of procollagen folding to couple it with the export machinery. 725 What controls the organelle size in the context of intracellular trafficking? There has been a lot 726 of work on what set the size of organisms, the size of tissues in an organism, and the size of 727 cells in a tissue. However there has been relatively less work on the question of what sets the 728 size of organelles relative to the cell. Extensive cargo transfer while trafficking bulky cargoes 729 such as collagens leads to large amounts of membrane being transferred from organelle to or-730 ganelle. To maintain organellar homeostasis, loss of membrane from a compartment has to be 731 concomitantly compensated by membrane acquisition from the biosynthetic pathway or by traf-732 ficking from other organelles; the arrival and departure of membrane at each compartment has 733 to be efficiently balanced. How is this homeostatic balance controlled? Changes in membrane 734 tension have been described to affect rates of exocytosis and endocytosis at the plasma mem- ficking by membrane tension is more challenging to study experimentally and hence still re-739 mains poorly understood. We propose that TANGO1 serves as a hub in the ER to connect 740 different organelles for the intracellular traffic by controlling the tension homeostasis and reg-741 ulating the membrane flux balance between these organelles. 742 743 In summary, we proposed a theoretical mechanical model that explains how TANGO1 mole-744 cules form functional rings at ERES, and how these TANGO1 rings assemble the machinery 745 required to form a large transport intermediate commensurate to the size of procollagens. We 746 envision that our hypotheses and the predictions of our model will open up new lines of exper-747 imental research to help understand how COPII coats organize together with proteins of the 748 TANGO1 family to allow for the export of folded procollagen out of the ER. 749

751 DETAILED DESCRIPTION OF THE PHYSICAL MODEL OF TANGO1 RING 752
FORMATION 753 TANGO1 filaments are described by their physical length, ? , which is proportional to the 754 number of protein monomers forming the filament; and by their persistence length, o = 755 ? / ™ , -where ? is the filament bending rigidity and ™ is the thermal energy, equal to the 756 Boltzmann constant times the absolute temperature (Doi and Edwards, 1986)-, which describes 757 how stiff the filament is. As long as the filament length is not much larger than the persistence 758 length, the bending energy of the TANGO1 filament can be expressed as rf\i = 759  TANGO1 fam-778 ily, however, for the sake of simplicity, we consider them all to be equivalent to each other. We 779 will only need to take into account this energy term when considering interactions between 780 neighboring rings, which involve a partial breakage of otherwise closed filaments (see Appen-781 dix 1). 782 783 Second, the effect of COPII polymerization on the ER membrane has two contributions on the 784 total free energy of the system: the first one is through the line tension, 3 , of a COPII-coated 785 membrane patch; and the second one is associated to the chemical potential of COPII area of )*)+ = 5 , then the total boundary length is = 2 . The free energy term con-797 tributed by the chemical potential of polymerization is 798 799 which describes, by classical nucleation theory, a minimum ERES size, ‡[\ = ª / 3 , above 802 which the polymerizing domain is stable and can dynamically grow (Frolov et al., 2006). 803 804 And third, we need to include an extra energy term, oYf\ , which includes all the factors that 805 modulate the domain size distribution, including the aforesaid chemical potential of COPII 806 polymerization (Heinzer et al., 2008). This phenomenological free energy term, oYf\ , should 807 have a local minimum at certain domain size, 3 ( ), which could in principle change by the 808 presence of TANGO1 and hence depend on the wetting fraction, . For the sake of simplicity, 809 we will disregard this dependence, and consider 3 as a free parameter in our model. Hence, 810 we can approximately express this free energy as a phenomenological free energy term for The effects of other known players, such as the complex spatiotemporal dynamics of ERES 826 components, the recruitment of ERGIC53-positive membranes by TANGO1, and the recruit-827 ment of procollagen are implicitly considered through effective parameters of the model. Ad-828 ditionally, one should in principle also consider the translational free energy of the filament 829 components, which is larger for filaments wetting ERES than for free filaments. However, this 830 contribution is relatively minor compared to the rest of contributions to the free energy and 831 hence we disregard it in our formal analysis of the system free energy. 832 833 In total, the extensive free energy of the system, , is the addition of the different free energy 834 terms in Equations (M1-M4) Here we present the detailed description and derivation, as well as the mathematical formalism 845 of the analysis of the physical model of TANGO1-dependent transport intermediate formation 846 presented in the main text. Our model builds on a previously presented mechanical model for 847 clathrin-coated vesicle formation (Saleem et al., 2015), which we extended to allow for the 848 growth of larger transport intermediates by incorporating (i) the effects of TANGO1 rings on 849 COPII coats; (ii) the reduction of the membrane tension by the tethering and fusion of 850 ERGIC53-containing membranes; and (iii) an outward-directed force ( Figure S3A). 851 852 Analogously to the clathrin vesicle model by Saleem et al. (Saleem et al., 2015), we consider 853 that the free energy per unit area of coat polymerization onto the membrane, U , has a bipartite 854 contribution arising from the positive free energy of COPII binding to the membrane, U 3 , and 855 from the negative contribution of membrane deformation by bending, so U = U 3 − 2 where U is the surface area of the membrane covered by the COPII coat, and o is the pro-868 jected area of the carrier, that is, the area of the initially undeformed membrane under the carrier 869 ( Figure S3B). In contrast to our previous analysis of the two-dimensional scenario of TANGO1 870 ring formation, here we consider the bending of the membrane away from the initially flat 871 structure, and so we do not consider the phenomenological term of ERES size, Equation (M4), 872 but rather the free energy associated to coat polymerization, Equation (M6). 873 874 We also consider a line energy for the coat subunits laying at the edge of the polymerizing 875 structure. This line energy per unit area reads as 876 877 where ( ) = 3 − Δ is the line tension, consisting on the line tension of the bare coat, 3 , 880 and Δ is the line tension reduction associated with the TANGO1-filement wetting; and = 881 2 is the length of the carrier edge, associated to the opening radius at the base of the carrier, 882 ( Figure S3B). 883 884 Next, we consider the effect of the membrane tension. We consider that the membrane is ini-885 tially under a certain tension, 3 , and it can get a local decrease in tension, Δ , by the fusion of 886 incoming ERGIC53-containing membranes. Hence, the actual membrane tension at a given 887 moment is = 3 − Δ . We can estimate that Δ =° )*±L² , where ° is the stretching 888 coefficient of the membrane, and )*±L² is the surface area of each of the ERGIC53-889 containing vesicles that fuse to the budding site (Sens and Turner, 2006). Hence, the tension 890 associated free energy per unit area reads, 891 892 where ‡ is the surface area of the entire membrane after deformation. 895 896 Next, we consider the contribution of the TANGO1 filament into the free energy of the system. 897 Analogously to our discussion for the free energy of coat binding to the membrane, Equation 898 (M6), we can write this free energy per unit area as 899 900 Finally, the mechanical work performed by the outward-directed force, , is also included in 913 the free energy of the system, as 914 915 where ℎ is the length of the carrier (Figure S3B). At this stage, we disregard the effects of the 918 growth-shrinkage dynamics of the polymerizing COPII lattice, as included in our formal anal-919 ysis of TANGO1 ring size through the phenomenological term in the free energy, Equation 920 (M4). Hence, the total free energy of the carrier per unit area, U , is the sum of all these contri-921 butions Equations (M6-10) Figure S3B, panels (i) to (iii), respectively). These shapes will allow us to cal-932 culate as a function of the carrier morphology the geometric parameters that enter in Equation  933 (5), namely, the area of the coat, U , the area of the membrane, ‡ , the projected area, o , and 934 the length of the coat rim, (Saleem et al., 2015). A convenient quantity to parametrize the 935 shape of the carrier is the height of the carrier, ℎ, which we will use in a dimensionless manner 936 by normalizing it to the diameter of the spherical bud, = ℎ/2 . 937 938 (i) Shallow bud. For a shallow bud (Figure S4B(i)), which corresponds to buds smaller than a 939 hemisphere, we can write that U = ‡ = 2 5 (1 − cos ), where 0 < < /2 is the open-940 ing angle of the bud (see Figure S3B(i)). In addition, o = 5 = 5 sin 5 ; and ℎ = 941 (1 − cos ). Expressing these quantities as a function of the shape parameter, , we obtain 942 943 U = ‡ = 4 5 ∶ < (ii) Deep bud. For a deep bud (Figure S3B(ii)), which corresponds to buds larger than a hemi-948 sphere, we can write that U = 2 5 (1 − cos ), where /2 < < . In addition, ‡ = 949 5 (1 + (1 − cos ) 5 ); o = 5 ; and ℎ = (1 − cos ). Expressing these quantities as a 950 function of the shape parameter, , which in this case ranges between ½< < 1, we obtain 951 952 plete bud with opening angle 0 < < , connected via a narrow connection with complete 959 buds ( Figure S3B(iii)). Here, we can write that U = 2 5 [2 + (1 − cos )], where 0 < 960 < and ≥ 1 . In addition, ‡ = 5 [4 + 1 + (1 − cos ) 5 ] ; o = 5 ; and ℎ = 961 (2 + 1 − cos ). Expressing these quantities as a function of the shape parameter, , we 962 obtain 963 964  The angle is then optimized as that corresponding to the minimal energy of the fused ring 1098 configuration with respect to the isolated ring configuration (Figure S4D). 1099 1100 We computed the energy barrier required to be overcome to allow ring fusion, Δ IÄ°[©\ (Ap-1101 pendix 1- Figure 1C). Our results indicate that a decrease in the interaction energy between 1102 TANGO1 filaments and COPII subunits leads to lower fusion energy barriers (Appendix 1-1103 Figure 1C) and hence more efficient ring fusion, as experimentally observed (Raote et al., 1104 2018). We also computed whether the overall fusion process is energetically favorable or not, 1105 which indicates whether the fused configuration can be even formed de novo before circular 1106 rings are fully assembled, indicating that negative filament spontaneous curvatures promote the 1107 stabilization of the fused ring configuration since they stabilize association of the concave face 1108 of the filament with COPII subunits (Figure S4G). Altogether, the results of our theoretical 1109 model show that the closer the system is to the wetting-dewetting transition, the more feasible 1110 it is to observe spontaneous formation of fused TANGO1 rings (Appendix 1-Figure 1D) To compute the preferred ERES size independently of TANGO1 interaction, we consider the 1342 situation of complete dewetting, = 0, which allows us to simplify Equation ( whereas the free energy loss associated with bending the filament upon wetting (Equation 1363 (M1)) is Δ rf\i,efXX =