After the examination of thylakoids in the light of evolution and ontogenesis, the following evidence has emerged: a) grana are, probably, not cylindrical structures of membranous discs; b) the formation of parallel-aligned thylakoids seems to be an intermediate stage before stacking.

The structure of pea LHCII trimer and its charge distribution on the stromal surface has been presented by Standfuss and colleagues,¹¹ who proposed that electrostatic attractive forces between LHCII of opposite membranes play a major role in the cohesion of grana. However, a single PSII/LHCII complex can interact electrostatically with many other neighboring PSII/LHCII complexes at any direction. The following interactions are distinguished: i) transverse interactions between PSII/LHCII - PSII/LHCII of the same thylakoid across the lumen. These interactions tether the opposite membranes, are responsible for the flat shape and the rigidity of thylakoids and monitor the width of the lumen (Figure 1A); ii) transverse interactions between PSII/LHCII - PSII/LHCII of adjacent thylakoids across the stroma. These interactions are responsible for the parallel appearance of adjacent thylakoids and regulate the distance between them. (Figure 1B); iii) lateral interactions between PSII/LHCII - PSII/LHCII within the plane of one membrane. These interactions are responsible for the lateral movement of the complexes and the formation of aggregates. (Figure 1C).

Any model concerning thylakoid stacking should explain all the different patterns of thylakoid conformation observed during evolution, i.e., from the parallel arrangement of lamellae, to the close associations of parallel membranes forming bundles and finally, to the formation of grana. It should also explain their different aspects during photomorphogenesis, i.e., from the parallel alignment, to the presence of appressions and finally, to the sudden appearance of stacks. In addition, it should involve a mechanism with the participation of LHCII and the implication of the physicochemical forces and this mechanism must be rapid and reversible.

In the model presented in this work, the protagonistic roles are given to electrostatic forces between the PSII/LHCII complexes.¹¹ Constant roles are distributed to electrostatic repulsion between thylakoids carrying net negative charges on their surfaces⁷ and between negatively charged PSI/LHCI complexes.¹⁰ Fission and fusion reactions, hydration repulsion⁷ and steric hindrance⁷ also play significant roles. The torque is the guest star appearing at the final act of thylakoid stacking.

At low salt medium, as already reported by Izawa and Good,⁵ thylakoids form a continuous network of parallel membranes (Figure 2A). Negatively charged membrane carboxyl groups²¹ and negatively charged PSI/LHCI complexes¹⁰ keep the adjacent thylakoids far apart, due to repulsive electrostatic forces (Figure 2, step 1). Furthermore a uniform lateral distribution of all complexes inside the membrane is also established.¹⁰

The addition of cations causes electrostatic screening on the surface of thylakoids.²¹ So the adjacent thylakoids approach each other in a parallel way. The PSII/LHCII complexes present both positively and negatively charged domains as proposed by Standfuss and colleagues.¹¹ Due to the balanced interactions between the repulsive forces exerted by the negatively charged PSI/LHCI complexes¹⁰ and the attractive electrostatic forces by the PSII/LHCII complexes, the adjacent membranes form bunches of parallel thylakoids (Figure 2, step 2).

Besides the transversal interactions, lateral attractive electrostatic forces are exerted between the PSII/LHCII complexes within the plane of one membrane. The PSII/LHCII complexes can move laterally only when the electrostatic forces that are exerted between them, are strong enough to overcome the resistance of the fluid mosaic. These forces increase either with the decrease of the distance between the complexes, or with the increase of their electric charges. The latter increase as the PSII unit size increases, while the distance between them decreases as their number increases during greening. Both conditions are accomplished in mature chloroplasts of angiosperms. Indeed Kohorn and Yakir²² have demonstrated that the major protein component of LHCII can migrate through thylakoids. Due to these attractive forces exerted between the neighboring PSII/LHCII complexes, the latter approach each other and form circular aggregates in both the upper and lower membranes of the same thylakoid. Such circular aggregates have been observed by Kouril and colleagues.²³ During the lateral diffusion of the PSII/LHCII complexes, some of them may be obstructed by other membrane proteins and may remain entrapped in stroma thylakoids. The negatively charged PSI/LHCI complexes and ATP synthase, which contain extended projections into the aqueous stroma, are excluded from the interior of the aggregates by steric hindrance.⁷ As the aggregates keep growing, they expand the membrane, forming discoid patterns with simultaneous sliming of both ends of the thylakoid. Stronger attractive forces are also exerted transversely between opposing complexes across the lumen, forcing the lumen to contract even more (Figure 2, step 3).

Attractive forces are also exerted on the complexes of all the parallel adjacent thylakoids, forcing the complexes to diffuse laterally and form large discoid aggregates. As the aggregates grow, the distance between the parallel-superimposed thylakoids decreases and the attractive electrostatic forces between the circular aggregates of the adjacent membranes become increasingly stronger, forcing them to come closer. As the neighboring underlying and overlying thylakoids approach each other, the membrane regions remaining outside the aggregates, at both ends of the thylakoid, exert increasingly greater repulsive forces, due to the negatively charged PSI/LHCI complexes. Therefore, a torque is formed, causing the superimposed thylakoids to rotate around the vertical axis of the circular aggregates until a static equilibrium is obtained in a lower energy level (Figure 2, step 4). Indeed, Chow and colleagues¹⁰ suggested that stroma thylakoids carry a higher net negative electrical charge than do stacked thylakoids. Hydration repulsion⁷ at the partition gap allows the rotation and the simultaneous sliding of the stacked membranes, permitting the alignment of the stacks. Under the rotation, stroma thylakoids interconnecting neighboring grana are disrupted. Very early Izawa and Good⁵ underlined such fragmentations. Indeed, these membranes become weakened, due to the dramatic decrease of PSII/LHCII complexes, which tether the two membranes across the lumen. After the formation of stacks, thylakoids may fuse with each other, establishing again a continuous membranous system. This assumption is supported by the presence of a protein that mediates fusion and affects thylakoid organization in Arabidopsis chloroplasts.²⁴

During fusion, bifurcations may be created between the disrupted membranes. Finally, the whole chloroplast obtains an ellipsoid shape as already reported by Izawa and Good.⁵

When stacked thylakoids (Figure 2, step 4) are washed and re-suspended in low salt solution, the PSII/LHCII complexes become negatively charged and segregate laterally, due to repulsive forces (Figure 2, step 3). As a result, the adjacent thylakoids are repelled away, due to their negative charges. The distance between them is increased (Figure 2, step 2) and their parallel arrangement reappears. Finally, a random distribution of all the complexes is established¹⁰ (Figure 2, step 1) and the whole chloroplast looks swollen in agreement with the observations by Izawa and Good (Figure 2A).⁵

Given that this model does not follow strict rules, it can explain all the irregularities observed. Thylakoids can rotate randomly either clockwise or anticlockwise. The aggregates can be formed anywhere inside the thylakoid, in the middle or at the edges, giving rise to different configurations. When formed in the middle, the stroma lamellae seem to cross the stack perpendicularly; when formed at the edges, the membranes seem to run from one level of the granum to a different level on the opposite side. In vivo, the stacks are formed by combination of these patterns. Even the number of appressed membranes per granum stack is not predetermined. This model accepts, as already proposed by Horton and colleagues²⁵ that the aggregation of the light harvesting complexes in vivo is induced by protonation of the lumen.

The rotational model is a new model that attempts to elucidate the mechanism of stacking, focusing on the evolutionary and ontogenetic continuity of thylakoids. Thus, it rejects the generally accepted idea that grana consist of cylindrical structures of superimposed discoid membranes. In this model, the principal role is given to attractive electrostatic forces exerted between the PSII/LHCII complexes. The main steps for stacking are, firstly, the close approach of the parallel thylakoids and the formation of bundles, secondly, the lateral migration of the PSII/LHCII complexes, thirdly, the formation of circular aggregates and fourthly, the rotation of thylakoids around the vertical axis of the aggregates. In accordance with these steps, the following observations should be mentioned. The formation of long parallel bundles and their coexistence with stacks in phaseolus vulgaris chloroplasts transferred to continuous light after prolonged preexposure to periodic light,²⁰ advocates that the formation of bundles is the first step. Kohorn and Yakir²² have demonstrated that the radiolabeled major protein component of LHCII can migrate through thylakoid membranes, showing that the second step (the lateral migration) is possible. The third step, the formation of circular aggregates, has been demonstrated by Kouril and colleagues,²³ who observed superimposed circular aggregates consisting of PSII/LHCII complexes in granal thylakoid membranes from spinach, using cryo electron tomography. The forth step, i.e., the rotation of thylakoids, which is a fast process, cannot be observed directly using electron microscopy. However, the consequences of the rotation, namely the extensive breakage of the membranes, have been reported by Izawa and Good,⁵ who noticed a considerable breakage of lamellae between grana. A prerequisite for rotation to occur is that unstacked thylakoids should be negatively charged relatively to stacked membranes. Indeed it has been proposed that stroma thylakoids carry a higher negative surface charge density than the surfaces in contact with the stacked membranes.¹⁰ This model predicts that the width of each superimposed thylakoid sheet is about 300-600 nm. As the thylakoids rotate they form a granum, which seems to have a phenomenal diameter of 300-600 nm.

Some experimental data can also be explained using the rotational model. It has been reported that plants grown in PSI light have a greater PSII/PSI ratio and a greater abundance of LHCII complexes, while they exhibit a greater extent of granal stacking and smaller non-appressed membrane domains, as compared to plants grown in PSII light.¹⁰ The greater extent of stacking was mainly attributed to a greater contribution to Van der Waals attraction, due to the increased amount of PSII light harvesting complexes. It was also attributed to the lateral segregation of PSI into the non-appressed membrane domains, because of the excess of negative charges it carries on the stroma-facing side of the membrane.¹⁰ According to the rotational model the greater extent of stacking can be explained by the presence of greater amounts of LHCII complexes, thus the distances between the complexes diminish and the attractive forces become strong enough to cause the migration of the complexes and the formation of aggregates and stacks by rotation.

When the chlorophyll-b-deficient CD3 mutants of wheat are grown under low to moderate light, they exhibit significant amount of stacking with more extensive arrays of parallel stroma thylakoids. Under high light conditions, very few membrane stacks were produced, with most of the thylakoids organized in long parallel membrane sheets.²⁶ According to the present model, thylakoids in high light have a parallel configuration, because the amount of PSII/LHCII complexes is not sufficient to provoke their lateral aggregation. On the contrary, when grown in low light, the number of PSII/LHCII complexes exceeds the required threshold and they migrate laterally, they form aggregates and finally, they rotate to generate stacks.

Izawa and Good⁵ found out that spinach chloroplasts isolated in low-salt media lose their grana and look considerably swollen; upon addition of cations, grana reappear and the chloroplasts shrink. Similar observations have been reported with intact bean and pea chloroplasts that were incubated with increasing concentrations of MgCl₂ and examined with confocal laser scanning microscopy.²⁷ It has also been noticed that in leaf samples of illuminated plants, chloroplasts have an ellipsoidal shape and their thylakoids show the well-known differentiation into unstacked stroma and stacked grana thylakoid membranes. In contrast, in leaf samples of plants after 3 or 7,5 hours of darkness, grana stacks are disintegrated to a large extent and chloroplasts look swollen.²⁸ If grana were vertical cylindrical stacks of membranous disks, one would expect the opposite to occur: in the presence of grana chloroplasts would have a swollen shape, while in their absence an ellipsoidal one. The present model denies the existence of cylindrical piles of membranous discs and gives a logical explanation to the different configurations of chloroplasts (Figures 2A,B).

A prerequisite for thylakoid stacking models is membrane continuity enclosing a single lumenal space. This model suggests that, during the rotation of thylakoids, disruptions may occur especially when unstacked thylakoids are directly converted into stacked by addition of cations in vitro. As already mentioned, such disruptions have been reported very early by Izawa and Good.⁵ However, this model advocates that in vivo, the fission of thylakoids occurring during the rotation may be accompanied by subsequent fusion of the membranes, which thus become continuous. The presence of a protein that mediates fusion and affects thylakoid organization in Arabidopsis chloroplasts strongly supports this admission.²⁴