Textbooks will tell you that as soon as the nucleus starts to divide, the Golgi sacs fragment into hundreds of tiny membrane-walled pockets called vesicles. These are then distributed at random between the two daughter cells; once the new cells start forming, the vesicles in each cell simply fuse together to make a new Golgi apparatus.

But this theory has been challenged by recent experiments. Kristien J. M. Zaal, of the National Institutes of Health, Bethesda, Maryland, and his colleagues, fire the latest salvo in the journal Cell1.

They present persuasive evidence that the Golgi membranes do indeed fragment at the start of nuclear division, but from then on the scenario is quite different from the one usually proposed. They find that the Golgi vesicles appear to be completely reabsorbed into the 'endoplasmic reticulum' or ER, another of the cell's internal membrane systems, only to re-emerge in the two halves of the dividing cell once cell division is well advanced.

The controversy over what happens to the Golgi at cell division is an extension of an on-going debate about the dynamic relationship between the Golgi and the ER as they carry out their normal business.

The conventional view is that the Golgi apparatus is structurally a quite distinct and stable organelle (as cellular organs are known). In this view, traffic from the ER to the Golgi is almost entirely one way; small vesicles budding off the surfaces of the ER move outwards and deliver their contents to the Golgi apparatus by fusing with its membranes. Traffic in the other direction is minimal and is limited to the return, in vesicles, of proteins that should have remained in the ER but accidentally escaped.

The alternative view is that there is extensive two-way traffic between the Golgi apparatus and the ER, with Golgi membranes continually cycling backwards and forwards between the ER and the Golgi. In this scenario, what happens at cell division, when both the Golgi and the ER shut down their functions temporarily, is that the backwards recycling of Golgi membranes to the ER continues, but forward movement stops entirely.

It resumes only when cell division is well advanced and the cell has nearly divided into two. Golgi vesicles again start budding off from the ER, which has in the meantime also split into two, and a new Golgi apparatus starts to form in each new cell.

Which of these two opposing ideas is correct is still hard to decide. Commenting on the topic in the same issue of Cell, Michael Roth of the University of Texas Southwestern Medical Center at Dallas, Texas, concludes that the two may even be "polar views of a process that in fact incorporates the essential elements of both"2.