Our Scientists

The endoplasmic reticulum (ER) is a continuous membrane system that consists of the nuclear envelope and the peripheral ER. Part of the ER, called rough ER, contains membrane-bound ribosomes that synthesize secretory and membrane proteins; regions devoid of ribosomes and involved in synthesizing lipids or inactivating toxic molecules are called smooth ER. Morphologically, the ER can also be divided into sheets, which include the nuclear envelope and perinuclear regions, and a network of interconnected tubules. The peripheral ER is very dynamic (Figure 1), with tubules continuously forming and disappearing and sheets rearranging. In mammals, the tubular ER network extends throughout the cell, whereas in yeast and plants it is located close to the plasma membrane and is referred to as cortical ER. The relative amounts of ER sheets and tubules vary greatly among cell types. Defects in ER morphology cause growth defects in budding yeast, decreased embryonic survival in the roundworm Caenorhabditis elegans, and a neurodegenerative disease known as hereditary spastic paraplegia in humans, suggesting that vital cellular processes rely on the morphological integrity of the ER.

Figure 1: COS-7 cells are transfected with ER-GFP, an artificial ER protein with signal sequence and ER retention signal, and visualized for 0.94 min by time-lapse confocal microscopy. Frames were taken every 0.57 s, and the video is shown at five frames per second. Bar, 5 µm. Video by Shaoyu Lin and Junjie Hu

How organelles are shaped is a fundamental question in cell biology. The ER, with its morphologically distinct domains, offers a unique system to study this problem. Recently, a major breakthrough in the field was achieved with the identification of two families of integral membrane proteins that shape ER tubules: the reticulons and DP1/Yop1p. The discovery was based on an in vitro ER formation assay developed in the Rapoport lab. These proteins are conserved in all eukaryotic cells, localize specifically to the tubular ER, and form homo- and hetero-oligomers. In the absence of the reticulons and Yop1p, the peripheral ER in yeast is converted from tubules to sheets.

We have combined biochemical and cell biology methods to address how the ER tubules are generated by the reticulons and DP1/Yop1p. When Yop1p or Rtn1p was purified and reconstituted into proteoliposomes, the membranes were deformed into tubular structures, suggesting that these proteins are sufficient to generate membrane tubules. Reconstitutions performed with various components reveal that tubule formation occurs with different lipids; requires essentially only the central portion of the protein, including its two long hydrophobic segments; and is prevented by mutations that affect tubule formation in vivo. We also found that the concentration of reticulons and DP1/Yop1p determines the diameter of the tubules and that these proteins form oligomers in the plane of the membrane. Thus, we propose that the reticulons and DP1/Yop1p utilize at least two mechanisms to generate ER tubules: (1) the two long hydrophobic segments would form a double-hairpin structure in the lipid bilayer, which would take the shape of a "wedge," generating the high membrane curvature that one sees in cross sections of ER tubules; and (2) the proteins would form arc-like oligomers around the tubules, which would serve as "scaffolds" to mold the membrane into tubules. My lab is trying different assays to test these hypotheses.

Once the tubules are formed, the next question is how they are connected into a network. Recent work from our group and others has provided evidence that membrane-bound GTPases are required for ER network formation—specifically, tubule branching. We showed that the atlastins (ATLs), a family of dynamin-like GTPases, interact with the reticulons and DP1 and are localized mostly to the tubular ER. Depletion or overexpression of the ATLs changes ER network morphology, and anti-ATL antibodies inhibit ER network formation in vitro. Most convincingly, purified full-length Drosophila ATL mediates liposome fusion in vitro in a GTP-dependent manner. These results suggest that ATL is responsible for the homotypic fusion of ER membranes.

To further understand the mechanism of ATL-mediated membrane fusion, my lab took a structural approach. Crystal structures of the N-terminal cytosolic domain of human ATL1 (cytATL1) reveal a GTPase domain and a three-helix bundle connected by a linker region. Two different conformations of cytATL1 were observed (Figure 2). Biochemical data confirmed that GTP binding induces dimerization of two GTPase domains, and GTP hydrolysis and release of Pi trigger a relocation of the three-helix bundles. These motions appear to be critical for the fusion activity of ATL, as they may allow the associated transmembrane segments to pull the membranes together so that they can fuse. We also showed that the C-terminal cytosolic tail of ATL (ATL-CT) plays an important role.

It is known that mutations in human ATL1 cause a common form of hereditary spastic paraplegia (HSP), a neurological disorder characterized by progressive spasticity and weakness of the lower limbs due to a length-dependent abnormality of corticospinal axons. When these mutations were mapped into the structures, two hot spots were identified: one at the dimer interface of the GTPase domain and the other on top of the helix bundle (Figure). Both would be expected to affect fusion in our model. Thus, HSP could be caused by ER fusion defects, and our results offer explanations at a molecular level.

ATL is well conserved among higher eukaryotes. In organisms that lack ATL, we have identified another class of membrane-bound GTPases as likely functional orthologs of the ATLs. They are called Sey1p in the yeast Saccharomyces cerevisiae and RHD3 in the plant Arabidopsis thaliana. They share the same signature motif and membrane topology. Sey1p interacts genetically and physically with the reticulons and Yop1p and cooperates with them in maintaining ER morphology. Mutations in RHD3 cause branching defects in the ER network and affect root hair growth in plants. Another member of the dynamin family, mitofusin (MFN) in mammals and Fzo1p in yeast, also mediates homotypic membrane fusion; it has the same membrane topology as the ATLs and fuses outer mitochondrial membranes. However, whether MFN uses the same GTP cycle-related dimerization and conformational changes to achieve membrane fusion remains to be elucidated. Our current research program involves studying the mechanism of membrane fusion mediated by these membrane-bound, dynamin-like GTPases in a systematic way and identifying additional factors for maintaining and regulating ER morphology. Our findings will not only provide new insight into the basic cell biology of membrane shaping and remodeling but also will elucidate pathogenic mechanisms, as these proteins are linked to several neurodegenerative diseases.

Grants from the National Basic Research Program of China and the National Natural Science Foundation of China provided partial support for these projects.

As of January 17, 2012