A Note on Three-Dimensional Models of Higher-Plant Thylakoid Networks

Ever since the electron microscope became commercially available in 1939, plant biologists have exploited this powerful tool to elucidate the intricate structure of higher-plant thylakoid networks to correlate structure with function and chromatic adaptability as well as to gain insight into the biogenesis and evolution of these membranes. During the following decades, several models have been proposed to describe the three-dimensional (3D) organization of chloroplast thylakoid membranes, in particular the spatial relationship between the two major morphological components: the appressed thylakoid regions, called the grana, and the stroma lamellae.

In a Perspective in this issue of The Plant Cell, Mustárdy et al. (2008a) present a model, entitled the “quasihelical model of the granum-stroma assembly,” for the architecture of higher-plant thylakoid membranes. To reinforce their model, they make use of and reinterpret data we published in this journal three years ago (Shimoni et al., 2005). Here, we offer some comments on the model they propose and describe why we believe our model accounts for the experimental data in a better way.

The helical model for thylakoid network organization was formulated by Paolillo (1970) and derived from observations originally made by von Wettstein (1959), Heslop-Harrison (1963), and Wehrmeyer (1964). The model subsequently was modified by Brangeon and Mustárdy (1979; see also Mustárdy, 1996) and by Mustárdy and Garab (2003). In its latest form, the model suggests a bipartite structure consisting of a cylindrical granum body, made of discs piled one on top of the other, around which the stroma lamellae are wound as right-handed helices. The grana connect to each other through the stroma lamella helices, which are tilted ∼20° with respect to the grana stacks and make multiple contacts with successive layers in the grana through apparent slits located at the rim of the stacks. Mustárdy et al. (2008a) present what is essentially this same model, albeit noting that it requires some revisions.

Originally, the helical model was derived from thin-section and serial-section electron microscopy (EM) analyses and later supported by scanning electron microscopy of sonicated chloroplasts (Mustárdy and Janossy, 1979). Thin-section micrographs provide two-dimensional projections of the specimen; hence, information about the spatial relationships between its constituents is inevitably masked by superposition. Serial-section EM allows the visualization of large objects in 3D, but it too bears the above limitation. Moreover, the resolution of this method in the depth dimension (z axis) cannot exceed twice the thickness of the section (e.g., 140 nm for 70-nm-thick sections) (McEwen and Marko, 1999, 2001). The data obtained from the scanning electron microscopy analysis (Mustárdy and Janossy, 1979) should be treated very cautiously as the sonicated samples likely suffered from fragmentation or deformation artifacts.

The aforementioned limitations can be overcome using electron microscope tomography (EMT), which currently offers the best means to extract high-resolution 3D information on cellular structures and organelles. We (Shimoni et al., 2005) employed EMT to determine the 3D architecture of higher-plant thylakoid networks; more recently, Mustárdy et al. (2008b) also used this technique toward the same end. However, sample preparation in these two works differs significantly. In our study, we imaged thylakoids within their native environment, namely, within leaves, whereas Mustárdy et al. (2008b) used isolated thylakoids. We agree with Mustárdy et al. (2008a) that thylakoids are quite robust and are therefore likely to remain relatively intact during isolation. Nevertheless, using leaf preparations or intact chloroplasts, where the system is kept under more native conditions, is more suitable for this type of analysis. Our samples were prepared by high-pressure freezing and freeze substitution, which preserve cellular structures with high fidelity (Lučić et al., 2005; Noske et al., 2008; Vanhecke et al., 2008). On the other hand, Mustárdy et al. (2008b) used conventional chemical fixation, which does not preserve cellular components as well as the above methods. Finally, it is difficult to assess the quality and integrity of samples in Mustárdy et al. (2008b) since none of their images provide an overall view of the thylakoids; shown are only high-magnification images of a single or at most a few grana (Mustárdy et al., 2008a, 2008b).

In addition to the differences in the choice of preparation methods, the two works also differ in data acquisition and, more importantly, in the subsequent analysis. We performed dual-axis EMT, which allows better resolution of elongated structures oriented at any angle in the x-y plane and also provides better resolution in the depth of the specimen than single-axis EMT used by Mustárdy et al. Thus, dual-axis EMT allows for better tracing of the connectivity of multiplanar systems, such as the thylakoid networks, than does single-axis EMT (Mastronarde et al., 1997). Following data acquisition and volume reconstruction, EMT proceeds by segmentation and rendering of the reconstructed objects, which is key to the visualization and analysis of the imaged structures. Although Mustárdy et al. (2008a, 2008b) remark that they generated model structures from their tomographic data, these models are not, in fact, presented in either of these reports. Instead, the analysis by Mustárdy et al. (2008a) seems to be a mere search for motifs that comply with the helical model, and despite noting that it requires important refinements, they only present a schematic model, which is identical to what they proposed previously (Mustárdy and Garab, 2003).

Segmentation and construction of 3D model structures from our tomographic data revealed that the granum-stroma assembly is formed by bifurcations of the stroma lamellar membranes into multiple parallel discs (Figure 1). The stromal membranes form wide, slightly undulating, lamellar sheets that intersect the granum body roughly perpendicular to the granum cylinder. Rather than winding around the grana, each stroma lamellar sheet enters and exits the granum body at approximately the same plane (Figures 1C and 1D). These features cannot be reconciled with a helical model wherein the stroma lamellae spiral around the grana and fuse to numerous granum layers at various planes. As shown in Figures 1A and 1B, adjacent granum layers are joined to each other not only through the stroma lamellae, as proposed by the helical model, but both via the bifurcations as well as through direct membrane bridges. The latter are formed by upward and downward bending of the granum discs that fuse to their neighbors at the edges (Figure 3 of Shimoni et al., 2005; also shown in Figure 1B of Mustárdy et al., 2008a).

Figure 1.

Reconstructed Structures of Granum-Stroma Assemblies.

Tomographic data were obtained from cryofixed, freeze-substituted, dark-adapted lettuce leaf preparations.

(A) The granum layers are contiguous with the stroma membranes that bifurcate (arrows) at the granum-stroma interface. Internal connections between adjacent layers are indicated by arrowheads.

(B) To better visualize the internal connections between neighboring layers in the granum (arrowhead), the structure shown in (A) was enlarged, clipped, and rotated by ∼40°, and the upper layer of the stack has been removed. In both panels, the structure was expanded along the z axis to provide a clear view of the interior of the granum.

(C) and (D) An ensemble of two grana interconnected by multiple stroma lamellae. The structure shown in (D) is a rotated view of the one shown in (C), with four of the lamellae removed. The stroma lamellae form wide, slightly undulating sheets that run parallel to each other and intersect the grana (surface-rendered gray objects) at an angle that is roughly perpendicular to the axis of the granum cylinder. The arrows shown in (C) define the plane of the layers inside the granum. All figures were adapted from Shimoni et al. (2005).

Forking or bifurcation of stroma lamellae into granum bodies had already been described in the 1950s (Hodge et al., 1955), as is invariably present in our images (Shimoni et al., 2005). Mustárdy and Garab (2003) stated that these forks or bifurcations were observed only rarely. Now, Mustárdy et al. (2008a) note that reexamination of their previous serial-section micrographs (Brangeon and Mustárdy, 1979), which were used to formulate the helical model, reveals that 15 to 20% of the granum-stroma lamellar junctions indeed appear to form forks. They attribute these forks to the low depth resolution of serial sectioning, which could cause neighboring slits to be superimposed to each other and mistakenly appear as forks. Although this argument may hold for data derived from serial-section analysis, it cannot account for our observations, which were derived from tomographic data. As can be seen in Figure 2, forks are visible in ∼2.6-nm-thick tomographic slices; hence, they are clearly not a result of convolution of adjacent slits. As discussed by Shimoni et al. (2005), when the tomographic volume is scanned in full, essentially all the stroma lamellae are seen to bifurcate into the grana. The reason it is impossible to visualize all the bifurcations of the stroma thylakoids within single cross-sectional views is that the bifurcations do not extend over the entire granum circumference, due to rotation and bending of the layers within the granum body.

Figure 2.

Forks or Bifurcations Are Not Convoluted Slits.

Shown are 2.6-nm-thick slices cut through the tomographic volume. Bifurcations of stroma lamellae into granum bodies are indicated by arrowheads. Bar = 65 nm.

Bifurcation of stroma thylakoids into grana is further supported by the observation that the ratio of stroma to granum thylakoids in dark-adapted chloroplasts, on which most analyses have been performed, is very close to 1:2. This ratio, which poses severe constraints on the structural parameters of the lamellar network within the framework of a helical model, was also observed by Paolillo (1970). Forks or bifurcations are invariably present in thin sections and tomographic images we obtained from both lettuce (Shimoni et al., 2005) and Arabidopsis (E. Shimoni, unpublished data). They have also been observed in tomograms of cyanobacterial thylakoid networks, suggesting that they emerged early in the evolution of thylakoid membranes, possibly as a means to attain connectivity in the lamellar system, and have since been preserved (Nevo et al., 2007).

A criticism raised by Mustárdy et al. (2008a) was that the contrast of our images, due to the use of high-pressure freezing and freeze substitution, does not suffice to discriminate between the thylakoid membranes and lumen. Accordingly, they claim that we could not achieve “a clear picture of the connecting regions,” as opposed to the contrast in their images of chemically fixed, stroma-depleted preparations. Application of high-pressure freezing and freeze substitution indeed renders the contrast in images lower than that obtained using chemical fixation; however, this is a price paid for preserving samples in a more native-like state. Importantly, low contrast does not necessarily entail low resolution: we could accurately trace thylakoids throughout the network: stroma lamellae, grana, and their interfaces alike (Shimoni et al., 2005). We further note that the lumen and membranes of thylakoids constitute a single entity; therefore, the above argument is in fact irrelevant to the ability to adequately probe the network. As a side note, we refer the reader to Shimoni et al. (2005) and Mustárdy et al. (2008a, 2008b) for comparison of the quality of the images obtained in the two studies.

The helical model requires that the stroma lamellae be tilted or inclined with respect to the granum axis to ∼20°. Indeed, in some cases the stroma lamellae are seen to be tilted (to various degrees) with respect to the granum cylinders that they intersect. Such variations in orientation are expected for large and deformable lamellar systems, such as thylakoid membranes. In most cases, however, the stroma lamellae are roughly perpendicular to the grana. This can be seen in the small ensemble shown in Figures 1C and 1D. It can also be seen by placing a pointer over a stroma lamellar sheet and inspecting the volume of the tomogram shown in Supplemental Movie 2 in Shimoni et al. (2005).

Finally, in Supplemental Figure 1 of their Perspective, Mustárdy et al. (2008a) show three slices that span 110 nm of the volume of a tomogram that we presented in Shimoni et al. (2005). In their analysis, they mark the shifts in the position of the stroma lamellae relative to the granum body when one travels through the volume. These shifts measure ∼20 nm in height, corresponding to about one granum layer. According to our model (shown in Figure 3 of Shimoni et al., 2005 and Figure 1 of Mustárdy et al., 2008a), bending of the granum discs at their edges and the intergranal thylakoid bridges account for these apparent shifts. However, for the helical model, given a tilt angle of 22° (Mustárdy et al., 2008a), one would instead expect a vertical shift of 45 nm. Moreover, the total thickness of the tomogram analyzed by Mustárdy et al. (2008a) was ∼170 nm. Scanning through the entire volume, we would expect to see even more extensive shifts (i.e., of ∼100 nm; corresponding to approximately five granum layers), were the stroma lamellae wound around the granum body. Such shifts are not observed either. More generally, for an average granum measuring 300 nm in diameter, the helical model predicts a pitch of ∼380 nm for each stroma lamella, corresponding to ∼19 granum layers, which will be seen as half-pitch shifts on either side of a cross-sectioned granum. Even when one considers variations in tilt angle and granum diameter, the shifts for spiraling stroma lamellae are still expected to be extensive. Such extensive shifts have not been recorded.

To date, there is no direct structural evidence supporting a helical model of thylakoid networks; what exist are observations showing occasional tilting or shifting of the stroma lamellae with respect to the grana. Given the size and flexibility of thylakoid networks, such tilts or shifts are quite expected and should not be used as indicators and certainly not as proof for such a model, which imposes severe topological constraints on an entire lamellar system. Forks or bifurcations have been observed, while slits are only inferred. Likewise, the well-established observations that the major photosynthetic complexes are segregated and that membrane stacking is promoted by granal-residing PSII-LHCII or LHCII arrays, which the authors cite in support of their model, in fact have no relevance to the winding of stroma lamellae around the grana. Our model (Shimoni et al., 2005), which is based on structures derived directly from tomograms of well-preserved samples, while undoubtedly in need of further improvement and refinement, is simpler and accounts for all structural and functional requirements expected from thylakoid networks of higher-plant chloroplasts. In the future, application of all-cryo EMT together with studies of network biogenesis will allow gaining insight into the architectural design of the thylakoid network in even finer detail.