Rho, Rho, Rho Your Cells

[talk script for the presentation]

The tile of my presentation is actually the first line of a nursery rhyme, which I've very slightly adapted for my talk today. It goes a little something like this:

Rho, rho, rho your cells
Give them U & P
Throw them in a nanopore
Look at all the genes

I'll get into that shortly, but first I want to talk a little bit about the mitochondrial genome.

Mitochondrial transcription

Here's a diagram of the mitochondrial genome, showing the location of genes in yellow and transfer RNAs in blue. Most of the genes and tRNAs are encoded on mitochondria in the same direction, but ND6 and a few tRNAs are encoded in the opposite direction (which I've indicated here as darker colours).

The current understanding of how the mitochondria is transcribed into genes is that the entire mitochondrial DNA sequence is transcribed as a unit, which is then chopped up and processed into individual genes, and eventually translated into proteins. If the life of a gene transcript were constant, then we might expect that those processed genes had similar expression. An expression experiment carried out on \emph{one} mitochondrial gene should give very similar results to that carried out on another gene.

Mitochondrial transcription - theoretical

Something a little bit like this, where I'm showing a profile of the mitochondrial genome with the location on the X axis and the amount of detected RNA sequence on the Y axis. There are gaps between the genes where the transfer RNA are removed prior to gene translation. But that expression pattern is just something I whipped up in an image editor a couple of days ago; it's not what you actually see when looking at expression across the mitochondrial genome.

Mitochondrial transcription (GL261)

It looks a little more like this. We did a cDNA sequencing experiment with a GL261 mouse brain tumour cell line last year, and I looked at the expression of all the processed mitochondrial genes. What we got was uneven coverage for genes along the mitochondrial genome. This result was not completely surprising to me. I'd seen a similar pattern in dendritic cells when I was looking at Illumina sequencing data for Franca's group.

But it surprised Mike a little. That expression pattern doesn't fit what would be expected, suggesting that there's some other level of control of mitochondrial gene expression. Mike's group has been investigating the transfer of mitochondrial DNA, and it's fairly
important for them to have a good understanding of the effect that mitochondrial DNA has on a cell.

Mitochondrial Duplication

So, we've started thinking of ways to try to work out what's going on. One thing that I'm aware of is that the mitochondrial genome is copied many times over throughout the nuclear genome, almost in its entirety in some places. One crazy explanation for the expression difference for mitochondrial genes is that these other copied genes are being transcribed and expressed. So there's potentially a situation where the expression that we observe in the mitochondrial genome is actually a function of two components:

Equation 1

Namely the nuclear expression of mitochondrial genes, and the mitochondrial expression of mitochondrial genes.

If that were happening, it would be important to find some way to exclude this background signal when investigating mitochondrial DNA transfer, as it might be confusing our attempts to work out what's going on.

I don't like this equation; it's too complex. What we need is some well-developed tool that can be used to make this equation a bit simpler.

Equation 2

Well, as it happens, the Berridge group have already demonstrated methods by which they can remove mitochondrial DNA from cells. They've even got a publication from it, and a bit of funding to continue their research.

So... by excluding the mitochondrial component of this equation, it would allow us to work out the nuclear component of mitochondrial protein expression...

Equation 3

and with a bit of shuffling also allow us to work out the mitochondrial component of expression.

Making rhos of cells

So, the Berridge group have this established toolkit for removing mitochondrial DNA from cells. They have stressed to me that it's not mitochondria that are being removed, just their DNA.

From what I've learnt from group meetings, the way to remove mitochondrial DNA from a cell line is to stop cells from producing mitochondrial DNA in the first place.

Cells can't survive without mitochondria, but they can survive without mitochondrial DNA, particularly if you feed them uridine & pyruvate.

They're still not completely happy, and telling the difference between a zombie cell and a partially-functioning rho-0 cell typically involves many months of waiting, and hoping that cell populations will start to expand.

Due to a number of issues removing all the mitochondria from GL261 cells, and keeping it removed, we switched over to 4T1 breast cancer cell lines for this experiment. So, we rho'd the cells, and fed them uridine and pyruvate to keep them alive. But before we could throw them in a nanopore, the RNA from the cells needed to be prepared for sequencing.

Nanopore Sequencing

There's a lot of information here, but I guess the message I want to give you around this is that nanopore sequencing is fast, cheap, and versatile, and it's going to get a lot cheaper a few months from now. We can use it for a whole bunch of things, but what I'm going to be talking about today is how it can be used for cDNA sequencing.

Strand-switching cDNA Generation

The sequencing template was prepared by extracting RNA from our 4T1 cells (both rho-0 and WildType), then adding a polyT primer. A reverse transcription enzyme extends from the polyT sequence through to the start of the gene, then leaves a little extra bit on the end. We can use this little extra bit as an anchor for the attachment of a primer for replicating the forward strand of the sequence, a process known as template-switching PCR. The really nice thing about template-switching PCR is that an entire polyA-tailed gene transcript can be converted into cDNA without knowing anything about its sequence.

There's a little niggle in that a RNA sequence without a polyA tail won't be targeted for sequencing. This provides us with an easy way to get rid of the bath-water of ribosomal RNA, but there are also a few micro RNAs and a lot of other non-coding RNAs that will be
ignored as well.

Barcoding & Amplification

During amplification in subsequent PCR reactions, additional primers are attached to the ends of the template that include barcodes for sample identification and binding sites for sequencing adapters. After the amplification has finished and unincorporated primers have been washed away, the adapters are added to the mix and a room-temperature ligation-free reaction is carried out that bonds the adapters to the prepared DNA template.

Nanopore flow cell

This hybrid DNA-adapter complex is the thing that is then carefully dropped onto the sequencing matrix of a MinION flow cell. The DNA is drawn towards sequencing pores with the help of a mild electric current. The DNA-adapter complex engages with that sequencing pore, and the MinION then records the changes in electric current as the sequence is ratcheted through one of 512 pores at the slow speed of 450 bases per second.

Nanopore Raw Signal

What you're looking at here is the electrical signal trace that was produced from a single adapter-bound cDNA template. This sequence took about 5 seconds to go through the nanopore. Once it has come through a sequencer and been stored as a sequence on a computer, we call it a read. The electrical trace gives a warts-and-all view of the DNA as it
moves through the pore. It can tell you about adapter sequences and polyA tails (or in this case a polyT head), but, also, it's got information about things we don't understand yet. If there's a modified base that somehow slipped through the PCR amplification steps, then that will show up in the electrical signal trace. This trace also shows, in this case, when the pore has stalled near the end of the sequence.

Nanopore Called Sequence

This signal is then converted into a more familiar DNA sequence, converting squiggly lines into A, C, G, and \& T. Here's the DNA sequence created by the signal in the previous slide, which is about 1,500 bases long. Because of all the things that influence the electrical signal, this base-called read is a little bit innacurate, but that's the price you pay for being able to detect things you didn't know you didn't know about. If you squint, you might be able to see a stretch of lots of Ts at the start of the read; that's a polyT head that tells me this read is a copy of the reverse-complement of the original RNA sequence.

Nanopore Annotated Sequence

We can use super-fast text-searching programs to look for specific sequences within other sequences; we call this mapping. In this case, I used a program called LAST to look for matches to the nanopore adapter sequences within the read that was produced by the
basecaller. I'm able to pick up fragments from all the adapters, which is good, and I can see from this that there are overlaps between the barcoding sequences and the strand-switch sequences.

So, instead of starting from a whole-transcriptome strand-switch protocol, we could have designed our own site-specific primers that have this region as an overhang, and then use the PCR barcoding kit for sample preparation prior to sequencing. This would allow us to do away with the need for a polyA tail, and see, for example, if we could find unprocessed mitochondrial transcripts in a sample of RNA.

On the other end of the sequence, it's a little interesting that there seems to have been a bit of clipping going on at the end of the gene, where it's skipped out a reasonable chunk of the strand-switching adapter. This might be related to the stall that is apparent in the electrical signal trace, but I haven't looked at that yet.

Just as an aside, I'd like to point out that the total length of the adapter sequences that were added to the input cDNA is about 200 bases, so any DNA sequencer that sequenced 100 (or even 125) bases from each side with these adapters would have trouble getting anything useful out of this sequence at all. The long reads that you get out of a nanopore device give you a lot of rope to play around with.

Mitochondrial Transcription

I've used another text-searching program called minimap2 to match up the million or so reads from our sequencing run to the mouse genome. Some of those reads matched up with the mitochondrial genome. Because reads included polyA or polyT sequences, together with their associated adapter, I was able to work out the direction in which a transcript was encoded on the genome.

So mitochondrial genes are expressed in the 4T1 wildtype cells that we prepared. There's a bit of reverse-strand expression happening, most of which has been observed and  reported by others before.

But I'm sure what you're all wondering is what we see in the rho-0 cells; in other words, those cells without mitochondrial DNA....

Mitochondrial Transcription (rho-0)

The answer, is nothing. No expression on the mitochondrial genome at all. So, going back to our equation...

Equation 1

We can rule out nuclear expression of mitochondrial genes.

Equation 2

At least, we can rule it out in cells that have no functional mitochondrial DNA. So, for now, we'll assume that the pattern of expression we see on the mitochondrial genome is real, and it's time to start thinking about what else could be going on to regulate the abundance of different mitochondrial genes.

Blank slide

Okay, but one obvious concern is that maybe the reason we didn't see any expression in the rho-0 cells is that something went wrong with the sample prep. Maybe we put water into the PCR reaction instead of RNA.

Well, apart from Carole and Olivier carefully demonstrating that that wasn't the case with agarose gels, I had my own evidence on the computer side in that the rest of the  transcriptome was clearly being expressed as well.

Acknowledgements

This is about where the song ends for us at the moment. We need to do some more replicates of this just to be absolutely sure about the lack of expression, and maybe look at other rho-0 cells from other cell lines. READ LESS