New Method Maps the Fatty Landscape of the Cell’s Protein Factory

The endoplasmic reticulum is the cell’s busiest workshop, a sprawling network of folded sheets and narrow tubules where nearly every new membrane protein is inserted, folded, and checked for quality. Yet for all its importance, the exact lipid recipe of this organelle in human cells has remained surprisingly murky.

The technique, called mammalian MemPrep, adapts a workflow originally developed for yeast. The researchers grew HEK293T cells engineered to produce tagged versions of two ER proteins: SEC61β, part of the protein-translocation machinery found across the entire ER network, and REEP5, a curvature-stabilizing protein concentrated in ER tubules. Each tag consists of a 3xFLAG epitope, a protease cleavage site, and a MYC epitope, arranged so that the tagged vesicles can later be snipped free from their capture beads. The cells were lysed gently with a Dounce homogenizer in a hypertonic buffer, avoiding detergents that would strip away membrane identity.

After lysis, a series of centrifugation steps separated the goods from the clutter. A spin at 1000 times gravity removed nuclei and unbroken cells; a second at 10,000 times gravity pulled down mitochondria, a notorious contaminant in ER preparations; and a final ultracentrifugation at 100,000 times gravity collected crude microsomes enriched for ER-derived membranes. The pellets were then briefly sonicated to break larger ER fragments into small vesicles, which were captured on magnetic beads decorated with anti-FLAG antibodies. Harsh washes containing 0.6 molar urea removed nonspecifically bound material, and the clean vesicles were released by cutting the tag with purified HRV3C protease. The whole procedure takes roughly sixteen hours from harvesting the cells to freezing the isolate.

Purity was verified at every stage. Immunoblotting showed that ER markers such as REEP5, Calnexin, and the calcium pump SERCA2 were strongly enriched in the final isolates, while markers of the plasma membrane, mitochondria, nucleus, endosomes, and cytosol were depleted to undetectable levels. Quantitative proteomics went further: the REEP5 preparation detected 6060 proteins including 437 annotated ER proteins, and the SEC61β preparation quantified 5191 proteins including 365 ER proteins, with ER-annotated proteins enriched far above proteins assigned to other organelles. Several apparently non-ER proteins that turned up highly enriched turned out on inspection to be genuine ER residents mislabeled in databases, including the reticulophagy receptor FAM134B and the SNARE protein BET1.

The trade-off for this purity is yield. Differential centrifugation alone loses roughly ninety percent of ER membranes, and after immunoisolation only about 0.2 percent of the total cellular ER lipid material survives in the final sample. The authors also caution that mild overexpression of the bait proteins, mechanical disruption of the network, and the possibility that some sub-organellar architecture is lost during lysis all remain limitations, even though the bait constructs showed minimal impact on the cellular proteome and lipidome in control experiments.

Comparing the two isolates revealed a clear proteomic split. The REEP5 preparation was selectively enriched for known tubule residents such as Atlastin 1 through 3, Lunapark, Extended Synaptotagmins, and REEP1, 3, 4, and 5. The SEC61β preparation, by contrast, carried ribosomal proteins, translocon components including SEC61α and SEC63, the insertase GET3, and the sheet-stabilizing protein CLIMP63. In other words, the method genuinely separates ER subdomains at the level of proteins, preserving at least part of the organelle’s internal organization through the isolation procedure.

The lipids told a different story. Quantitative shotgun lipidomics showed that the two isolates, despite their distinct proteomes, featured essentially identical lipid compositions. Phosphatidylcholine dominated at more than 50 mole percent, cholesterol sat low at roughly 12 mole percent, and monounsaturated fatty acyl chains were enriched at the expense of polyunsaturated ones. Compared with whole-cell lysates, the ER membranes were dramatically depleted of phosphatidylserine, cardiolipin, and cholesteryl esters, lipids that mark the plasma membrane, inner mitochondrial membrane, and lipid droplets respectively, further underscoring the cleanliness of the preparation. Ether-linked phosphatidylcholine was enriched in the ER while ether-linked phosphatidylethanolamine was depleted.

The acyl-chain analysis added finer detail. ER glycerophospholipids favored short fatty acyl chains and shunned long ones, and lipids with one or two double bonds were enriched over those with three or more. Extrapolating from model membrane systems, the authors argue this recipe should produce loose lipid packing and a highly compressible, relatively thin bilayer, exactly the environment needed for insertases that locally distort the membrane to thread new proteins into it. Previous work has shown that raising ER cholesterol blocks membrane protein insertion and that lipid saturation triggers ER stress, so keeping the bilayer soft appears to be a matter of survival for the organelle.

One curious finding concerned sphingolipids. Ceramides isolated from the ER carried more hydroxylations than ceramides in the whole-cell lysate, a chemical signature not seen for sphingomyelin, whose species distribution matched that of whole cells. The authors note that this added polarity echoes the increased polarity they observed in ER transmembrane helices, though they stress that the functional meaning of ceramide hydroxylation remains to be worked out.

That protein-lipid echo extends further. Using a new bioinformatic tool called HelixHarbor, the team analyzed the transmembrane helices of 82 single-pass ER proteins detected in their experiments and compared them with previously characterized sets of 90 Golgi and 273 plasma membrane helices. ER helices were shorter, bulkier, and significantly less hydrophobic than their plasma membrane counterparts, with p values as low as 1.59 times ten to the minus fourteen for the hydrophobicity difference. ER and Golgi helices were statistically indistinguishable from each other. The authors interpret this concordance, together with fewer positively charged juxtamembrane residues in ER proteins matching the ER’s lower density of anionic lipids, as evidence that membrane proteins and their surrounding lipids have co-evolved under shared biophysical constraints. The result is a coherent picture of an organelle whose tubules and sheets may differ in their protein workforces but apparently share one common, carefully tuned lipid fabric.

Subject of Research: Lipid and protein composition of endoplasmic reticulum membranes isolated from HEK293T cells using a new immunoisolation method

Article Title: Mammalian MemPrep establishes the lipid composition of ER membranes in HEK293T cells

Article References: Jain, A., von der Malsburg, A., Alsayyah, C., Götz, C., Elmofty, M., Reinhard, J., Haberkant, P., Helms, V., Lorent, J. H., & Ernst, R. (2026). Mammalian MemPrep establishes the lipid composition of ER membranes in HEK293T cells. EMBO Reports. https://doi.org/10.1038/s44319-026-00952-9

Image Credits: AI Generated

DOI: 10.1038/s44319-026-00952-9

Keywords: endoplasmic reticulum, MemPrep, lipidomics, proteomics, HEK293T cells, phosphatidylcholine, cholesterol, transmembrane helices, ER tubules, membrane protein insertion, immunoisolation, EMBO Reports