Separating cells is hard

I write this entry to accompany my short talk at Woodstock.bio meeting ( #physiologicalirrelevantconference ) .

A few years ago we published a paper in PNAS in which we showed that full-length mRNAs transfer between mammalian cells via a unique type of structure called membrane nanotubes, or tunneling nanotubes (TNTs). This work was started at Rob Singer’s lab, continued at the Gerst lab and in collaboration with Arjun Raj.

I wrote a “behind the scenes” post, detailing how that paper came to be, and some of the problems I had along the way.

I next published a method paper, which also included some new information – in particular that the transferred mRNA is encapsulated in an unknown protein shell. I wrote a “behind the paper” post at the Springer Nature blogs. There, I told about all the problems I had just because a simple change of the cell fixation conditions of my FISH protocol.

The problem is depicted on the right side of my Woodstock.bio slide:

Very briefly – because the regular FISH protocol leads to TNTs breakage and loss, I decided to increase TNTs stability by adding glutaraldehyde to the fixation buffer. This led to a four-fold increase in TNT preservation. But the transferred mRNA disappeared! It took me a very long time to figure out what’s going on there and partially solve this – at the expense of TNTs’ stability again.  I still have hopes to find a fixative that will preserve the TNTs without affecting the FISH quality.

The left side of the slide depicts our grad student’s greatest achievement – something we’ve been trying to get at over the past six (6!) years. The idea is very simple – co-culture human and mouse cells. After some time, separate then to pure human or mouse cell populations and send for RNA-seq. This should reveal the entire transferome – which human mRNAs are found in the mouse cells and vice versa. As a control, we have a mix of human & mouse cells which were cultured separately, mixed and immediately separated in parallel to the co-culture.

The issue is that we need very high purity. This is because we estimated the amount of transferred mRNA as 1% or less of the endogenous. So if we have 1% donor cell contamination, it will obscure the transferred mRNAs.

For about 2-3 years, I tried to separate the cells with flow cytometry, using various labeling strategies and conditions. But I never managed to get a clear signal of our positive control (MS2-labeled mouse beta-actin mRNA) in co-culture over mix. Then Sandipan Dasgupta joined our lab and instead of FACS sorting, he used affinity purification with magnetic beads to sort the cells. It seemed to be going fairly well. So much so that we also designed an in vivo experiment in mice. We then sent our samples to sequencing only to find out that the sequencing facility had made some mistakes, or there was another problem and all our samples were either contaminated with mouse RNA, or just mixed somehow. That facility closed (we were last in queue ) so there was no way to solve it. But, we also learnt that we probably would not have had enough coverage anyway.

So, Sandi repeated the (in vitro) experiment in order to collect new samples for RNA seq – but we noticed, based on more quality control experiments we did, that the separation was not good enough for us. Although the mouse cells were very pure (99.9%), the human cells always had a small level of mouse cells (98.5% purity of the human cells). Since our expected signal is about 1-2% of the mRNAs being transferred, we could barely see a signal in co-culture compared to mix (1.3-fold).

So, Sandi worked really hard, playing with the conditions until he solved it, and got consistent 99.9% purity of the human cells – just a few months ago. The qRT-PCR result in the slide shows 4-5 fold more human beta-actin mRNA in mouse cells in co-culture compared to mix (we have similar results for the mouse beta-actin mRNA in human cells). The samples were shipped for deep RNA seq (150 million reads per sample) and we are waiting for the results.

We also have more experiments going on – but these stories are for another time.

“Maybe we should open a falafel stand” is an actual text from Jeff when we discussed one Saturday evening on Whatsapp about all the problems we encounter in our experiments.