Intracellular membranes: Conserved mechanisms of formation and regulation throughout evolution

Membrane remodeling and phospholipid biosynthesis are normally tightly regulated to maintain the shape and function of cells. Indeed, different physiological mechanisms ensure a precise coordination between de novo phospholipid biosynthesis and modulation of membrane morphology. Interestingly, the overproduction of certain membrane proteins hijack these regulation networks, leading to the formation of impressive intracellular membrane structures in both prokaryotic and eukaryotic cells. The proteins triggering membrane proliferation share two major common features: 1) they promote the formation of highly curved membrane domains and 2) they lead to an enrichment in anionic, cone-shaped phospholipids (cardiolipin or phosphatidic acid) in the newly formed membranes. Taking into account the available examples of membrane proliferation upon protein overproduction, together with the latest biochemical, biophysical and structural data, we explore the relationship between protein synthesis and membrane biogenesis. We propose a mechanism for the formation of these non-physiological intracellular membranes that shares similarities with natural inner membrane structures found in α-proteobacteria, mitochondria and some viruses-infected cells, pointing towards a conserved feature through evolution. We hope that the information discussed in this review will Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 23 March 2020 doi:10.20944/preprints202003.0329.v1 © 2020 by the author(s). Distributed under a Creative Commons CC BY license. give a better grasp of the biophysical mechanisms behind physiological and induced intracellular membrane proliferation, inspiring new biotechnological applications.


Introduction
Membranes are lipidic films that define the boundaries of cells and organelles. They constitute both a permeability barrier for aqueous compounds and a major site of exchange between the interior and exterior of these cells and compartments. Therefore, they are essential for compartmentalizing the biochemical reactions that sustain life. Membranes are composed of lipids arranged as bilayers, together with proteins that can be either inserted in the lipid layer or peripherally associated to it. Membrane organization, as well as lipid and protein constituents, vary between organisms (eukaryotic cells, bacteria, virus), but also among species of certain organism (bacteria). Furthermore, membrane composition changes in response to various signals or environmental conditions resulting in three-dimensional rearrangements, or membrane remodeling events. These spatial rearrangements occur in all life forms and are essential for important physiologic processes, such as cell division and differentiation, organelle formation and intracellular trafficking. The exact mechanisms underlying membrane remodeling have mainly been deciphered in eukaryotic cells and are starting to be understood in prokaryotes. Importantly, the knowledge gained from studying membrane remodeling in physiological context has allowed the establishment of a few basic physico-chemical principles that appear to also apply to non-physiological membrane rearrangements, such as membrane proliferation occurring upon protein overproduction. complex membrane remodeling processes were thought to be an exclusive features of eukaryotic cells. Recently, it has been demonstrated that prokaryotic cells also undergo multiple membrane remodeling processes which are similarly controlled by specific proteins, analogous to the ones found in eukaryotic cells (Vega-Cabrera and Pardo-López 2017). Regardless of the different protein complexes involved for each organism, from a biophysical perspective, four basic molecular mechanisms have been described to remodel biological membrane, modifying their curvature (Figure 1) (Zimmerberg and Kozlov 2005  Those biophysical mechanisms are ubiquitous and involved in multiple physiological processes. For example, proteins with ENTH (Epsin NH2-Terminal Homology) domains in eukaryotes insert a wedged-shaped N-terminal amphipathic helix into the membrane, leading to curvature and deformation of the membrane (Figure 1a). ENTH-containing proteins are also involved in intrinsic curvature in its polymeric state and deform various artificial membranes in vitro (Kretschmer et al. 2019), led to a model in which the cytoplasmic membrane is pulled inward by Z-ring constriction during cytokinesis. However, it has also been suggested that FtsZ only serves as a scaffold onto which the peptidoglycan remodeling machinery assembles. In the latter case, it is the growth of the septal cell wall that pushes the membrane toward the center of the cell (den Blaauwen, Hamoen and Levin 2017). In addition, the actin-like ATPase FtsA, which interacts with phospholipids via its C-terminus and bridges FtsZ to the membrane, was shown to induce membrane rearrangement in vitro and vesicle formation upon overexpression in E. coli (Conti, Viola and Camberg 2018). It was thus proposed to facilitate membrane invagination by deforming the membrane at the septum site (Krupka et al. 2014;Conti, Viola and Camberg 2018). Whatever its exact mechanism of formation, the membrane curvature generated upon membrane invagination in turn participates in the recruitment of negative curvature-specific proteins such as DivIVA, which further binds other players of cell division and localizes them at the septum site. The final steps of bacterial cell division (fusion and fission of the membrane(s) leading to the separation of the two daughter cells) are not characterized yet and it is still unclear whether specific fusion/fission proteins complexes are necessary or if membrane fission occurs spontaneously as a consequence of membrane curvature and/or protein crowding (Snead et al. 2017;Steinkühler et al. 2020). A role for FtsA in this process has been proposed based on the occasional scission observed when FtsA was added to FtsZliposome in vitro (Osawa and Erickson 2013). However, this is inconsistent with the fact that et al. 2004) and accumulates in the mature spore (Kawai et al. 2006). Mutant strains producing only trace amounts of CL show delay in spore formation and produce reduced amounts of spores that are unable to germinate when placed back in favorable conditions (Kawai et al. 2006). CL enrichment might thus be important for the function of membrane proteins required for sporulation (e.g. FisB) or for their recruitment to specific regions of curvature. Membrane curvature-dependent localization has indeed been shown for B. subtilis SpoVM, which is necessary for spore maturation and localizes at the forespore surface by detecting positively curved membranes and inserting in them by an atypical amphipathic -helix (Ramamurthi et al. 2009;Gill et al. 2015).

Evolutive origin of intracellular organelles
Although prokaryotic cells have been historically claimed as organelle-free organisms, several examples of intracellular membrane-restricted compartments have now been identified. For instance, intracellular membrane structures are naturally present in α-proteobacteria, an evolutive ancestor of γ-proteobacteria (Gupta 2000), where they either increase the efficiency of the cell bioenergetic metabolism (anaerobic anoxygenic photosynthesis, nitrifying and/or methanotrophic bacteria, etc.) or provide an evolutive advantage (magnetosome) (Muñoz-Gómez et al. 2017).
Those membrane structures have been proposed as potential ancestors of mitochondria inner membrane cristae after the discovery of a common membrane remodeling protein: alphaMic60 and Mic60 in α-proteobacteria and mitochondria, respectively (Muñoz-Gómez et al. 2015). The growth of intracellular membrane structures in both α-proteobacteria and mitochondria requires the assembly of the photosynthetic or respiratory protein complexes, which are known to induce strong membrane curvature (Woronowicz et al. 2013;Horvath et al. 2015;Niederman 2016).
Mic60 and its analogue alphaMic60 are part of the protein complex that presumably bends the membrane and stabilizes the cristae junctions in mitochondria (or inner membrane invagination points in α-proteobacteria). Although modern γ-proteobacteria lack the gene encoding alphaMic60 (Huynen et al. 2016) and have lost the ability to physiologically produce inner liposome tubulation in vitro. Because a mamY Magnetospirillum magneticum mutant showed altered magnetosome size-distribution, MamY was first proposed to be involved in membrane constriction (Tanaka, Arakaki and Matsunaga 2010). However, MamY in vitro tubulation activity is specifically increased upon CL interaction, suggesting that MamY might recruit CL to the site of magnetosome formation to induce the formation of highly curved membranes (Tanaka et al. 2018). Still, overexpression of MamY in E. coli or M. magneticum did not alter cell membrane morphology, confirming that in vivo, other factors are certainly needed to trigger membrane curvature and vesicle formation (Tanaka, Arakaki and Matsunaga 2010). More recently, another study proposed that MamY represents a membrane positive curvature-sensing element and serves as a scaffold to properly align the chain of magnetosome parallel to the axis of the cell (Toro-Nahuelpan et al., 2019). The role of CL in this function was however not tested.
In some photosynthetic bacteria, intracytoplasmic vesicles called chromatophores contain pigments and light-harvesting proteins used to perform photosynthesis. Chromatophores function depend on the light-harvesting complexes 1 (LH1) and 2 (LH2) together with the reaction center (RC). These complexes, which are also directly implicated in chromatophore formation and shape determination, are thought to induce membrane curvature through a combination of wedging and scaffolding mechanism ( Figure 1). Indeed, the ability of these integral membrane proteins to bend and deform membranes depends on their capacity to oligomerize. The RC-LH1 complex, when monomeric, cannot bend membrane. However, RC-LH1 in complex with the small protein PufX forms dimers with the two monomers bent by a 146º angle (Qian, Bullough and Hunter 2008;Hsin et al. 2009;Tucker et al. 2010). In the absence of LH2, these dimers form tubular chromatophores in vivo (Chandler et al. 2008). LH2 is also sufficient to induce membrane curvature in R. sphaeroides. The protein forms hexagonally packed complexes, which are localized at high membrane curvature regions and, according to molecular dynamic simulation, could also induce membrane curvature (Chandler et al. 2009;Hsin et al. 2009;Scheuring et al. 2014). The combined action of LH2 and RC-LH1-PufX would thus allow for the formation of spherical shaped chromatophore.

Hijacking membrane remodeling: lessons learned from viral infection
In addition to the aforementioned membrane-remodeling physiological events, intracellular membranes can also be reshaped during infection by peculiar viruses able to usurp host lipid metabolism to create new compartments dedicated to their replication or replication organelles ( Figure 2). Viruses infecting a large variety of hosts, ranging from bacteria and unicellular eukaryotes to vegetal and animal cells, have been described that trigger this phenomenon  In addition, a correlation between membrane curvature and lipid biosynthesis during +RNA viral infection has been proposed (Miller and Krijnse-Locker 2008). Indeed, viral proteins modulating membrane curvature were also shown to promote the formation of membrane contact sites and the recruitment of host factors involved in lipid metabolism (van der Schaar et al. 2016), jn particular phosphatidylinositol-4-phosphate (PI4P) or phosphatidylethanolamine (PE) synthesis (Altan-Bonnet 2017). Interestingly, the accumulation of PIP4 or PE is often accompanied with an enrichment in sterol that might contribute to the stabilization of membrane curvature and is important for the replication of the virus ).
In addition to +RNA viruses, other viruses containing double-stranded RNA (Reoviruses) as well as DNA viruses (Poxvirus, Vaccina virus, African swine fever virus, Frog Virus 3 and Paramecium Bursaria Chlorella Virus, giant Mimivirus Acanthamoeba polyphaga) also induce massive host membrane rearrangements. Although less studied, those viruses also rely on the production of proteins modifying the curvature of the host membrane (Weisberg et al. 2017;Moss 2018, Mutsafi et al. 2013). The membrane-enveloped double stranded RNA bacteriophages from the cystoviridiae family, such as phage φ6, are the only known enveloped phages and are evolutionarily related to the +RNA eukaryotic virus picornavirus (Koonin, Dolja and Krupovic 2015). They also produce proteins capable of bending the inner membrane of their hosts (Gram negative bacteria) which are necessary for virus replication (McGraw, Mindich and Frangione 1986).
In summary, certain viruses hijack the lipid metabolism of their hosts using specific proteins that modify membrane curvature and host co-factors to alter the lipid composition of the membrane to favor their own replication. However, how those factors are related to de novo membrane biosynthesis and viral replication organelles assembly remains elusive and should be further investigated.

Inner membrane proliferation upon overproduction of some membrane proteins.
Overproduction of recombinant membrane proteins is usually difficult due to various limitations, including a shortage of membrane space needed to accommodate the produced proteins. In a few peculiar cases however, overproduction of membrane proteins, either in prokaryotic or eukaryotic cells, has revealed an unexpected and intriguing ability of cells to synthesize an excess of internal membranes. In fact, these newly synthesized inner membranes often contain large amounts of well-folded recombinant proteins, holding great promises for biotechnological applications. Since the pioneer observation of Weiner et al. (Weiner et al. 1984), only a few dozen membrane-associated proteins from prokaryotic (Table 1) (Weber et al. 1996) c) S. cerevisiae cell with the cytosol (cyt), nucleus (n) and the staked membranes "Karmellae" (k) around the nucleus (n) (left) and detail of those membranous structures (right) after 3-Hydroxy-3-methylglutaryl-CoA reductase (Hmg-CoA) overproduction (Wright et al. 1988). d) Vesicles formed in S. cerevisiae upon overproduction of poliovirus protein 2BC (Barco and Carrasco 1995).
From a morphological point of view, those inducible intracellular neo-membranes can be related to the bioenergetic compartments of α-proteobacteria and mitochondria (Arechaga 2013), and/or replication organelles of +RNA viruses (Miller and Krijnse-Locker 2008). Furthermore, similarly to those "natural" intracellular compartments most of the proteins triggering inducible intracellular neo-membranes (Table 1 and Table 2) also create zones with high membrane curvature (Jamin et al. 2018). For this reason, "natural" and "induced" intracellular membrane proliferation might share a more profound relation that goes beyond simple morphological resemblance.

Mechanisms of protein-induced membrane curvature
Modulation of membrane curvature is often at the midst of both physiological and induced membrane remodeling processes. The induced membrane proliferation upon protein overproduction has, however, the advantage of being decoupled from the cell physiological regulations. For this reason, we will examine the current knowledge on the mechanism of inner membrane proliferation, focusing on the influence of membrane curvature not only on membrane morphology but also on phospholipid biosynthesis. In particular, four important questions about inner membrane proliferation upon protein overproduction remain: 1) how can overproduced proteins induce the deformation of the inner membrane creating different morphologies; 2) what are the characteristics of the proteins triggering lipid biosynthesis and, thus, inner membrane proliferation; 3) how are protein overproduction and de novo phospholipids biosynthesis coordinated; and 4) can we find regulatory mechanisms conserved across evolution explaining internal membrane proliferation in both prokaryotic and eukaryotic cells.
In order to yield the observed morphologies (vesicles, tubules, stacks of flat membranes, etc.) listed in Table 1 and Table 2 and illustrated in Figure 3, proteins inducing inner membrane proliferation must modify membrane curvature by means of one (or the combination of several) of the general mechanisms previously proposed (molecular motors, supramolecular scaffolding, asymmetric membrane interaction, wedging, see Figure 1). However, the production of a pulling or pushing force will not be discussed in this section since none of the reviewed proteins is a molecular motor, nor a scaffolding protein interacting with any molecular motor.

Protein-protein supramolecular interactions
The construction of a 3D supramolecular scaffold via supramolecular interactions is an efficient way of controlling cell membrane curvature and is used by many proteins involved in membrane remodeling processes (e.g. endocytosis, fission, motility, membrane trafficking, etc.) (Simunovic et al. 2016). Most of the proteins inducing intracellular membrane proliferation (Table 1 and Table 2) have been described to form supramolecular assemblies around the lipid bilayer. The membrane curvature and, consequently, the inner membrane morphology observed in electronic microscopy will depend on the shape, nature and concentration of monomers constituting the supramolecular scaffold.
Heterologously expressed caveolin-1 in E. coli cells is perhaps the best-characterized example of how a single membrane protein can shape the morphology of the newly synthesized lipid bilayer. Caveolin-1 is a scaffolding protein involved on the formation of vesicles (caveolae) arising from the plasma membrane of eukaryotic cells (Parton 2018). The formation of heterologous-caveolae (h-caveolae) derived from E.coli inner membrane is linked to the assembly of caveolin-1 into a supramolecular cage (Walser et al. 2012). This cage contains around 160 caveolin-1 monomers and is similar in structure and size to eukaryotic caveolae.
The three membrane-interacting domains and the oligomerization domain of caveolin-1 are Preprints (www.preprints.org) | NOT PEER-REVIEWED | Posted: 23 March 2020 doi:10.20944/preprints202003.0329.v1 required for inner membrane proliferation (Ariotti et al. 2015). The formation of a regular and well-defined caveolin-1 scaffold seems to impose a strong local curvature on the cell membrane, causing the budding of vesicles coated with caveolin-1, and triggering the biosynthesis of phospholipids. As a consequence, monodisperse vesicles of the same size as those found in eukaryotic cells, accumulate in the E.coli cytosol.
The overproduction of phage PM2 protein P6 represents another instance of membrane proliferation presumably induced by a supramolecular cage (Männistö et al. 1999 producing onion-like vesicles instead of its characteristic replication organelles (Weber et al. 1996). The lack of other viral proteins or the differences in the nature and composition of phospholipids between eukaryotes and prokaryotes might explain this change in morphology.
The precise mechanism of membrane deformation by FMDV protein 3A is unknown, but it requires the central amphipathic helix of the protein, together with the two cytosolic N-and C-  Overexpression of E. coli fumarate reductase results in the formation of an array of densely packed lipid tubules in E. coli cytosol, that are severed from the inner membrane (Elmes, Scraba and Weiner 1986). These lipid tubules are stabilized by a scaffold of fumarate reductase packed in a regular helical configuration containing 10 proteins per helix turn (Weiner et al. 1984). Similar tubules are also observed when succinate dehydrogenase is overexpressed in E.coli (Maklashina, Berthold and Cecchini 1998). Probably, the mechanism of tubule stabilization is similar to that observed with fumarate reductase due the structural and functional similarities between these two enzymes (Ruprecht et al. 2009;Starbird et al. 2018). Although the supramolecular packing of succinate dehydrogenase has not been studied in depth, the authors observed different morphologies (tubules or vesicles) depending on the expression level of the protein (Maklashina, Berthold and Cecchini 1998). The supramolecular array of succinate dehydrogenase necessary to stabilize the tubule morphology might only be formed if the protein is produced at sufficient level. Thus, below a critical concentration, succinate dehydrogenase is still able to deform the membrane and yield vesicles but it is not capable of maintaining the tubule structure. The sn-Glycerol-3-P acyltransferase is another example of protein inducing membrane tubules formation upon overproduction (Wilkinson et al. 1986). The individual molecules of sn-Glycerol-3-P acyltransferase are arranged in dumbbell-shaped dimers, which are packed in a left-handed helix along the tubule axis (Wilkison et al. 1992). The association of six sn-Glycerol-3-P acyltransferase dimers completes a helix turn.
The whole FoF1 ATP synthase ( The overproduction of the serine chemotaxis receptor (Tsr) from E. coli also triggers inner membrane proliferation (Lefman et al. 2004). Tsr is a transmembrane protein with a periplasmic domain that binds small molecules (Tsr is specific to serine) and a cytoplasmic domain associated with the adaptor protein CheW and the kinase ChewA (Grebe and Stock 2004). In normal physiological conditions, cytoplasmic domains of adjacent Tsr form trimers of Tsr dimers, and self-assemble in two-dimensional clusters concentrated at the bacterial cell poles (Kim, Wang and Kim 2002). When overproduced, Tsr is also organized as trimeric assemblies of dimers (Lefman et al. 2004). However, because Tsr amounts are significantly increased, the two-dimensional clusters of Tsr can interact with each other creating a three-dimensional pseudo-hexagonal crystalline array that folds the inner membrane (Lefman et al. 2004). If this crystalline array is destroyed, e.g. by overproducing Tsr partners (ChewA and ChewW) at the same levels as Tsr, membrane proliferation is inhibited, even at high Tsr concentration in the membrane (Zhang et al. 2007). This result suggests that the high membrane curvature imposed by the crystalline array of Tsr is necessary to trigger phospholipid biosynthesis.
Membrane curvature induction by protein overproduction is not restricted to prokaryotic hosts and has also been observed in eukaryotic cells (Table 2). Unfortunately, structural data on the arrangement of the recombinant proteins in the newly synthesized inner membranes are lacking.

Insertion of wedge-shaped proteins into the membrane.
The insertion of wedge-shaped protein into the lipid bilayer can also modulate membrane curvature (Lee 2004). Most transmembrane domains of membrane proteins introduce a packing mismatch in the lipid bilayer, which alters the membrane curvature. A well-known example of this mechanism is the wedge-shaped bacteriorhodopsin, a light-driven proton pump expressed in archaebacteria under anaerobic conditions. Bacteriorhodopsin is found in highly curved, specialized membrane microdomains (purple membranes), in which it forms trimeric, hexagonal units packed in a 2D crystalline lattice together with archaeal lipids (Oesterhelt and Stoeckenius 1971;Henderson 1977). These lipids also contribute to the 2D crystalline packing as bacteriorhodopsin mutants with a constitutive wedge-shaped structure still need a specific lipidic environment to induce membrane curvature (Krebs and Isenbarger 2000;Rhinow and Hampp 2010;Yokoyama et al. 2014). Therefore, both phospholipid composition and protein tertiary 3D structure work together to modulate the membrane curvature in purple membranes.
It is remarkable how newly produced inner membranes of prokaryotes are enriched in coneshaped non-bilayer forming lipids (cardiolipin (CL) or lyso-phospholipids) (Table 1). Similarly to wedge-shaped proteins, those cone-shaped phospholipids can also modulate membrane curvature. Their effect is expected to be less important than those induced by proteins, but not  Table 1 and Table 2. Furthermore, tubule-shaped membranes are formed independently of the shape of the protein when the concentration of proteins is high enough to cover more than 40% of the membrane surface area, which explains the prevalence of tubular membrane structures upon protein overexpression. This study strongly suggests that the mechanisms involved in membrane curvature induction and, as a consequence, the E. coli maintains a constant ratio between zwitterionic phosphatidylethanolamine (PE), which accounts for about 75% wt. of total phospholipids, and anionic PG and CL, whose relative amounts depend on the physiological state (log-or stationary-phase) of the cells (Cronan, Jr. and Rock 2008). A feedback mechanism between the cross-regulated enzymes controlling the synthesis of PE and PG/CL (PssA and PgsA, respectively) maintains the homeostasis in phospholipid headgroup diversity (Figure 4). PssA is a monotopic membrane protein that acts as a sensor, detecting changes in relative phospholipid composition (PE vs PG/CL) in the lipid bilayer (Louie, Chen and Dowhan 1986;Satomi et al. 1996). It is active when associated with anionic phospholipids (PG and CL) and catalyzes the synthesis of PE. On the contrary, when anionic phospholipids become less available, PssA is deactivated, causing PgsA metabolic route to accelerate and to incrase the synthesis of PG and CL.
Besides the aforementioned enzymatic regulation, phospholipid homeostasis is also subject to  al. 2017), while intracellular membrane proliferation has only been observed with a dozen of specific membrane proteins. To determine whether proteins able to induce membrane proliferation display specific charges distribution, we calculated the surface electrostatic potential of the proteins listed in Table 1 for which a structure is available in the PDB. Those proteins present a positive electrostatic lobe located nearby the phospholipid polar heads in the cytosolic leaflet of the inner membrane ( Figure 5a). However, this feature is far from being  and transmembrane Fumarate reductase, PDB code 6AWF (Starbird et al. 2018)); b) proteins whose overproduction does not trigger membrane proliferation (MsbA, PDB code 6BPL (Mi et al. 2017;Oh et al. 2018) and G3P transporter, PDB code 5XJ9 (Huang et al. 2003)).

Formation of CL microdomains
An alternative explanation for the perturbation of phospholipid homeostasis in a reduced accessibility of certain type of phospholipids to the membrane homeostasis sensors, that could In this regard, the insertion of a membrane protein (or supramolecular complex of proteins) inducing a high local curvature would be needed to provide the necessary force to bend the membrane and induce CL clustering. This could in turn lead to an anionic phospholipids depletion in the non-curved zones of the membrane, which will subsequently be detected by the phospholipid homeostasis sensors as a signal to start lipid biosynthesis. Then, the newly synthesized lipid membranes would allow for more membrane protein insertion, thus, closing the cycle ( Figure 6). As previously discussed in section 2, all the proteins triggering inner - (Lefman et al. 2004;Zhang et al. 2007) sn-glycerol-3-P acyltransferase TM E. coli E. coli Tubules Helical (6 dimers per turn).
No changes. Dependent on phage heat shock protein (PspA). (Wilkinson et al. 1986;Wilkison and Bell 1988;Wilkison et al. 1992 Ariöz et al. 2013Ariöz et al. , 2014Ge et al. 2014) observations were expanded to HMG-CoA reductase isozyme (Hmg1) overproduced in ∆ire1 yeasts. In this case, membrane proliferation was achieved in complete independence from Ire1 and secretion of Kar2/Bip chaperone. Consequently, the authors concluded that, at least for Hmg1, membrane proliferation phenomena should be unrelated to UPR. However, the activation of UPR pathway seems somewhat advantageous for inner membrane proliferation, as the overproduction of the transcription factor Hac1 (the main product of UPR), using an external expression plasmid improved the production yield of membrane proteins and intracellular membranes (Guerfal et al. 2010;Vogl et al. 2014). Interestingly, overproduction of Hac1 alone changes the morphology of the ER membrane to a cubic phase and increases the Kar2/BiP chaperon levels (Guerfal et al.

Inositol regulation pathway and the importance of phosphatidic acid
Therefore, the existence of additional regulatory pathways to control phospholipid biosynthesis upon protein overproduction cannot be ruled out. In fact, intracellular membrane proliferation of ER in S. cerevisiae is associated with altered membrane trafficking. For example, overproduction of Sec12p blocks the ER-to-Golgi intracellular trafficking of S.cerevisiae and induces the formation of clusters of the chaperone Kar2/BiP, like in UPR pathway (Nishikawa, Hirata and Nakano 1994). Similar blockage of intracellular trafficking was also observed after overexpression of the poliovirus 2BC protein and the peroxisomal Pex15p protein, which both accumulate newly formed ER membranes (Barco and Carrasco 1995;Elgersma et al. 1997). Conversely, the overproduction of the canine RRp enhances the secretory pathways in S. cerevisiae (Becker et al. 1999). Of note, this altered intracellular trafficking is not a general scenario for all the protein-induced intracellular membrane proliferation in eukaryotes, as exemplified with Hmg1 (Nishikawa, Hirata and Nakano 1994). It is not clear whether these alterations of cellular trafficking are only a consequence of membrane proliferation or, on the contrary, contribute to inner membrane proliferation in eukaryotes. In any case, regardless of the precise mechanism, the biosynthesis of phospholipids also seems to be controlled by the membrane composition in eukaryotes.

Cardiolipin and phosphatidic acid membrane microdomains: a universal regulator mechanism for
phospholipid biosynthesis conserved through evolution?
PA seems to be a central actor in phospholipid biosynthesis regulation in eukaryotes (Carman and Han 2011).
For instance, increased PA cellular levels, either by a lack of PA degradation due to lower PA phosphatase activity or by an increase in PA concentration due to an overproduction of diacylglycerol kinases, leads to an expansion of the nuclear membrane (Santos-Rosa et al. 2005;Han, Siniossoglou and Carman 2007;Han et al. 2008). It should be noted that this membrane expansion in the absence of any membrane protein modulating membrane curvature leads to yeasts with an aberrantly large nucleus, without any organized membrane morphology (e.g. stacks of membranes, tubules or vesicles).  The two ionizable positions have been highlighted in blue in the 2D chemical representation.
As discussed before, proteins inducing membrane proliferation in both prokaryotic and eukaryotic cells are characterized by their ability to bend the membrane. At the same time, cone-shaped anionic PA and CL share a marked preference for negatively curved regions of the membrane. Therefore, the disturbance of phospholipid homeostasis via the creation of membrane microdomains enriched in PA or CL might be a central regulator of phospholipid metabolism conserved through evolution.

Membrane overproduction as a platform for biotechnological applications
Beyond the exotic observation of inner membrane proliferation in prokaryotes and eukaryotes, the phenomena of inducible membranes have interesting implications for biotechnology. The increased phospholipid amount due to membrane expansion could be useful in the field of biofuel production by fermentation. Modification of metabolic pathways aiming at diverting carbon fluxes towards the desired target compound has been tried and is far from being straightforward (Hollinshead, He and Tang 2014; Zhou et al. 2016). In this context, the overexpression of a protein triggering membrane proliferation could represent a simple, alternative strategy to redirect lipid metabolism and enhance biofuel production yield. In the same line, production of the P9 and P12 viral proteins have been proposed to increase the yield of useful hydrophobic active principles (Myhrvold, Polka and Silver 2016).
For structural biologists, membrane protein production still represents a major biochemical challenge (Zoonens and Miroux 2010). In addition, these production platforms could find applications in nanotechnology. Almost all cell types secrete nano-and micro-sized vesicles used for intercellular communication (Raposo and Stahl 2019). Granting control over the production and composition of those vesicles hold great promises in nanotechnology and nanomedicine (Busatto et al. 2020). In this regard, the protein-induced intracellular membrane proliferation could increase the vesicle production yield. For example, membrane proliferation upon overproduction of the b subunit of FoF1-ATP synthase has been recently used to prepare proteoliposomes (Royes et al. 2019). This method represents an attractive alternative to in vitro proteoliposomes reconstitution, alleviating several steps of protein extraction, purification and reconstitution in liposomes. In the same line, the preparation of bacteriaderived lipid vesicles presenting antigenic proteins from pathogens on their surface have been used for vaccine preparation (Farjadian et al. 2018). The production of chimera proteins containing a membrane-proliferation domain and an adequate antigen could dramatically improve vaccine safety and mass production. In any case, the mechanistic considerations revealed in this review can help to better understand membrane proliferation HMG-CoA S. cerevisiae S. pombe S. cerevisiae "karmellae" stacked membranes around nucleus ER Soluble domain not required. Transmembrane helix alone not sufficient. (Wright et al. 1988;Lum and Wright 1995;Profant et al. 1999) Cytochrome b5 R. norvegicus S. cerevisiae "karmellae" stacked membranes around nucleus ER Transmembrane domain disturbed by proline hinders membrane proliferation (Vergères et al. 1993) Cytochrome P450 C. maltosa S. cerevisiae "karmellae" stacked membranes around nucleus + Tubules ER Minimum domain 1-33: contains hydrophobic helix and charged residues flanking it. (Ohkuma et al. 1995;Kärgel et al. 1996;Menzel et al. 1996;Zimmer et al. 1997;Sandig et al. 1999) PMA2 (H + ATPase) S. cerevisiae S. cerevisiae Tubules ER - (Supply et al. 1993) RRp -180kDa C. lupus S. cerevisiae "karmellae" stacked membranes around nucleus ER RBS not required for membrane proliferation. Increase of secretory pathway. (Wanker et al. 1995;Becker et al. 1999 (Miller et al. 2003)