My cell biology papers

When I was an undergrad at Duke University, I worked in the lab of Dr. Sharyn Endow and ended up contributing¹ to two peer-reviewed cell biology papers. Here, for the benefit of my friends who are not cell biologists, I’d like to explain what those two papers reported.²

Scientific background

Motor proteins are molecules, present in nearly all eukaryotic cells, that convert chemical energy into mechanical work. Some of the most famous motor proteins are the kinesins, which are responsible for transport of cellular cargo and chromosome segregation during mitosis and meiosis.

Kinesin’s role in chromosome segregation means that it can be studied with the methods of classical genetics, such as by mating two organisms and observing the resulting offspring. One of the best organisms for performing such genetics experiments is Drosophila, the common fruit fly, thanks to its numerous offspring, well-mapped genome, and short generation time (~10-12 days).

Kinesins generally perform their functions by acting on microtubules, which are long filamentous polymers. During cargo transport, kinesins attach to the cargo and “walk” along microtubules. During chromosome segregation, kinesins act on the spindle, a structure composed of microtubules attached to the chromosomes.

Spindles occur in two different contexts: during mitosis, the division of a cell into two genetically identical daughter cells, and during meiosis, the process that produces reproductive cells for sexual reproduction.

The spindle can be studied with confocal fluorescence microscopy. A confocal microscope is one that uses a series of pinhole apertures and mirrors to only capture a subset of the light being emitted by the sample, enabling high-resolution 3D imaging. Fluorescence occurs when a molecule absorbs one frequency of light and emits another. By controlling the prevalence and localization of fluorescent molecules, researchers can control what the confocal fluorescence microscope “sees”.

In vitro means “in glass” and refers to experiments done in a controlled lab vessel like a test tube or Petri dish. In vivo means “in the living” and refers to experiments done inside live organisms and cells.

The day-to-day experience

One thing that may be of interest but is not described in the published papers is the day-to-day experience of doing some of the experiments.

Raising fruit flies

Raising flies is well understood but somewhat laborious.

Flies need food and shelter. We kept flies in vials containing homemade cornmeal agar fly food. We were careful to label each vial with the relevant genetic information for the flies inside, otherwise it might be impossible to tell simply from their external appearance. Stocks of each strain were maintained by periodically transferring adults to breed in new vials with new food.³

Moving flies between vials was straightforward but somewhat forceful. We first placed the source vial on ice, stunning the flies inside. We then removed the top cotton plug, inverted the source vial on top of the target vial, and shook the flies from the source vial into the target vial.

Performing a controlled mating (a “cross”) of flies was also straightforward. A male and a female fly would be left alone in a vial for a couple of days. Successful mating was indicated by the presence of eggs, embryos, larvae, and pupae, all clearly visible in the vial. Once the female had laid eggs, both adults were removed to avoid confusing them for offspring.

After mating, the flies were anesthetized and transferred to a “fly morgue” containing mineral oil.

Imaging oocytes

We often wanted to image individual egg cells, called oocytes. To do so, we first anesthetized a female fly on ice. Then, we dissected the fly, using fine tweezers under an optical microscope to remove the ovaries and isolate individual oocytes.

The oocytes were then placed on a glass slide and taken to a separate room with the laser confocal microscope. Acquiring a single high-resolution image could take tens of seconds as the confocal microscope “scanned” each pixel individually. In all, we would typically spend a couple of hours collecting numerous images, each showing different parts of the oocytes in different stages.

The KLP10A paper (2014)

KLP10A (“kinesin-like protein at cytological region 10A”) is a kinesin in Drosophila. It was first characterized by Rogers et al. (2004) who showed, through imaging and depletion experiments, that it localizes to the mitotic spindle and is required for correct spindle function.

Rogers et al. focused on KLP10A’s role in mitosis. Naturally, one subsequent line of inquiry sought to understand its function in meiosis. Zou et al. (2008) first successfully placed KLP10A in the Drosophila meiotic spindle. Radford et al. (2012) supplied the decisive loss-of-function genetic test, showing that the meiotic spindle becomes extraordinarily long when KLP10A levels are exogenously decreased.

Our KLP10A paper Do et al. (2014) builds on those two papers to further study KLP10A in meiosis. Specifically, we depleted KLP10A levels via two different mechanisms (RNAi and P-element loss-of-function) and reported reduced female fertility, elongated/mispositioned spindles, and normal microtubule growth rates. Our paper expanded on the existing literature in a couple of ways:

• First, it provided a mechanistic refinement to Radford et al., suggesting that the elongated spindles were due to reduced microtubule disassembly, not increased assembly.
• Second, it added microtubule growth rates (inferred from fluorescent measurements of EB1, a protein that associates with the growing ends of microtubules) as a new readout, linking KLP10A not just to static spindle morphology but also to spindle dynamics.
• Third, it contributed a partial-loss angle. Compared to the severe germline mutants in Radford et al., we reduced KLP10A levels less dramatically while still observing striking results.

The Ncd paper (2026)

Ncd is another kinesin in Drosophila. It was first reported in Endow et al. (1990) and was the earliest example of a kinesin motor protein that was shown to be related to chromosome segregation.

One of the ensuing lines of work sought to understand Ncd’s structure and force production. Hatsumi & Endow (1992) showed that Ncd affects spindle structure. Chandra et al. (1993) elucidated Ncd’s basic internal architecture by analyzing the function of some of its subregions in vitro, which Sablin et al. (1996) refined with a crystal structure of Ncd’s motor subunit. Furuta et al. (2013) successfully measured Ncd force production in vitro.

These papers gave us a reasonable understanding of how the different parts of Ncd work together to produce force, and even gave us direct in vitro measurements of Ncd’s force production. But could we also get in vivo measurements? Such measurements would add to our understanding of how Ncd behaves in its native cell environment (as opposed to the relatively artificial in vitro environment), but they are also difficult to gather as cells are very fast and crowded places.

One way to extract a signal from cells is by using fluorescence. A relatively precise measurement is possible because cells have low levels of background fluorescence, and confocal microscopes can precisely measure fluorescence levels. Grashoff et al. (2010) ingeniously exploited these properties to build a force biosensor suitable for use inside motor proteins. This biosensor is composed of two fluorescent molecules separated by a spring-like flexible linker protein. The microscope measures the level of fluorescent energy transfer between the two molecules, from which we can infer the distance between the fluorescent molecules. This distance can then be used to calculate the force that is acting upon the spring.

Our paper Do et al. (2026) inserted this force sensor into Ncd, a construct we called TsNcd (“tension sensor Ncd”). Because the force sensor we inserted into Ncd is very large⁴, it wasn’t clear if TsNcd would behave the same as Ncd. To see if this was the case, we compared TsNcd to native Ncd across many experiments: electron microscope experiments to compare their shapes, microtubule gliding tests to compare their in vitro function, fluorescence microscope experiments to compare their microscopic in vivo behavior, and Drosophila mating experiments to compare their macroscopic in vivo function.

TsNcd passed all these tests⁵, giving us the confidence to proceed with using it to make one of the first direct in vivo measurements of force across Ncd. From these measurements we reported several interesting results:

• Force across Ncd is greater inside the spindle than outside of it.
• Force across Ncd differs between assembling and mature spindles.
• The forces across Ncd regularly exceed the forces that it can produce itself.
• We can define different mechanical “states” for Ncd. Sometimes it acts actively as a motor; at other times it acts passively as a crosslinker.

Reflections

Looking back at the work I did over 10 years ago, a couple of thoughts come to mind.

• Cells are less inspectable than software. In software development, we have access to print statements to expose any state we care to know. If we need more power, we can use a debugger like gdb. These days, we even have debuggers like rr that let us step backwards in time. Cell biology has none of this.
• Different fields require different heuristics. One useful rule of thumb in software development is “You can understand everything if you keep trying.” In deep learning, this heuristic is less useful. Who can really say what the loss landscape looks like? But in cell biology, this heuristic can be actively misleading: the cell is often too complex to fully untangle.
• I’ve grown as a researcher. In writing this post, I realized that I didn’t have a good understanding of the scientific context in which I operated. Things turned out well because Dr. Endow kept us pointed in the right direction, but in hindsight this was clearly negligent on my part. These days, in my software engineering and machine learning work, I make sure to think about the broader context and whether the approach I’m currently pursuing is truly the best one.