Control DNA Experiments

A quick update since we last talked about how DNA controls were proving to be problematic: Things are working again (more or less). I went back top basics, remade most of my solution, ordered new DNA samples, and, most importantly, piranha-cleaned my chips. The first three experiments I did on basic 30-nm thick 5.4mm square chips I made myself, and I got DNA (2000bp, 2000bp, and 5000bp) through each one without too many issues. Whether it was Piranha or something else that made the difference is not clear, but it is a matter of superstition now so I will keep doing it for the time being.

An interesting outcome of these experiment was learning that sample loading technique in the flow cell makes a surprisingly large difference. In our old liquid cells (no flow), I would simply spike in a small amount of concentrated sample and mix it in-situ briefly, and diffusion would take care of mixing. In the flow cell, diffusion happens far too slowly, and sample must be pre-mixed before being injected. Doing so makes an event rate difference of a factor of 10-100, from <0.05Hz event rate to about 1Hz. It’s possible that some of my issues previously were due to this factor.

Finally, I ran a control experiment on a chip with the oxide pattern (2-micron holes with 3-micron spacing, no NPN yet) which I Piranha cleaned prior to use, and I was able to see 1000 bp DNA passing the pore. The signals were very odd-looking, but they are clearly DNA and they have a reasonable event rate. Experimental details:

 

The pore itself was 7-nm in diameter, assuming a 20-nm membrane thickness.

It was a little noisier than I would normally like for a DNA experiment, but since I am running pretty low on these I can’t be too choosy. Here is the baseline (-200mV applied, corresponds to negative bias on the oxide patterned side with the etch pit side grounded). Before adding DNA, we get a relatively high 1/f noise baseline, but no signals that could be mistaken for DNA:

After adding 1000bp DNA at an effective concentration of about 22nM, I get lots of events once I ramped the voltage up to -400mV (-200mV seems to be too low for a good event rate on membranes much thicker than 10-nm, at least so far).

 

This is where things start to get weird.

 

I am filtering at 100kHz. Normally, this means there should be a ~10 microsecond rise time for DNA events, where the signal, which should be a square step, gets distorted by the filter. You can read about it in detail here: http://pubs.acs.org/doi/abs/10.1021/nn405761y. Because these DNA molecules are very short, you expect to get a fair amount of attenuation from the filter and some reconstruction is needed to make sense of things, and I’m fine with that since I am really just looking for confirmation that DNA goes through, and I have that. However, the events look like this, close up:

 

The first thing to notice is that the effective rise time is a lot longer than it should be for this particular filter. In fact, where we expect something on the order of 10 microseconds for the rise time, MOSAIC extracts the following distributions for the rise times:

 

The peak at ~ -3 is what we expect to see given the filter settings. The higher peak is something I haven’t seen before. The long RC time suggests that something unusual is happening  – either the presence of the oxide spacer adds significant access resistance (unlikely, since the 2um holes would have negligible electrical resistance compared to the pore) or for some reason the effective bandwidth is reduced.

 

We can look at blockage depths of events when partitioned by RC constant. We expect a blockage of about 3nS for dsDNA in these conditions. Without any partitioning beyond limited RC constants to greater than 0.3 microseconds we get this:

We see a strong peak at 0.5nS and another peak at 3nS as we hoped. Looking only at RC constants between 0.3 and 10 microseconds we get only the first peak:

and looking at RC constants between 10 and 1000 microseconds we get the second peak:

The peak at 0.5 likely corresponds to collisions, where molecules get close enough to the pore to be seen but do not actually enter it. There are usually not nearly so many of those as we see here, suggesting that the oxide spacer might be causing some steric hindrance to DNA passage. But the odd thing is that we can say that molecules which fully translocate the pore have much higher rise times than the filter should cause. My first guess as to why is that the access resistance of the oxide well would be enough to register in the nanopore, since a gradually increasing blockage would potentially mimic an exponential rise time, and the RC constants in question are comparable to the entire translocation time, but some quick calculations show that R_well / R_pore < 0.00005, so this seems unlikely. Another option is that the oxide spacer somehow distorts the shape of the access region to produce this effect (something that could be simulated, if this becomes a trend with these devices). I don’t think it could be that the bandwidth is actually changed, since rise time comparable to the RC time would mean heavily distorted events which would not show the blockage level we are looking for.

However, if that’s not the case, then the usual fitting models might not apply to these oxide devices. I don’t think anyone has explored nanotopography effects on translocation dynamics before, so there might be some interesting work to be done in this area even without the NPN layer.

I am going to run a few more controls (next one planned with 5000bp DNA so that blockages will reach their full height and stay there for a while to facilitate fitting, on a 2um hole/3um space chip). I will also run DNA in the other direction next time to see if the odd behavior happens when DNA moves enters on the opposite side to the pattern. Then I will run a similar experiment on a chip with posts, to see if I can distinguish different nanotopographies from DNA signals. I still have the nanopore device that generated this device mounted, so I will see if I can clean it out and run DNA backwards in that one as well (after a few days, it may be fully clogged).

Experiments that actually have NPN on them will hopefully begin next week, once I iron out my membrane transfer issues, but I think there is enough interesting stuff to explore here that it’s worth spending some time to understand what the oxide layer is doing to translocation dynamics. Or maybe it will turn out to be a quirk of this particular pore. Stay tuned!