41J Blog

I wanted to knock together a quick visible demonstration of the nanopore/coulter counter sensing technique. This post documents a demonstrator using bits I had knocking around the house…

What is Ionic/Nanopore sensing?

(Ionic) Nanopores and coulter counters are devices that let you electrically detect the passage of small particles though a hole. The “hole” (pore, aperture) sits between two chambers which are filled with a conductive fluid. A voltage is then passed between the chambers and a current flows through the pore. Anything that passes through the pore blocks this current flow. So by measuring the current, you can detect the passage of particles through the pore. When a particle is in the pore, the current can’t flow as well as usual. So you get a short, drop in current as the particle blocks the pore.

Big particles block the aperture more than small ones. And you can therefore also get size information. This principle is used in the coulter counter to measure particle size distributions (for example to get white blood cell counts, or understand the distribution of particle in paints, or printer toner).

Building a macropore!

The basic sensing method described above works at a range of scales. All the way from millimeters down to nanometers. To get a feel for the basic principle lets build a macropore!

I used a bottle cap, which I made a ~2mm hole in (I used an old soldering iron, but a drill would be better). The toothpicks in the side (completely blocking small holes), are to hold the “pore” in place later.

Next we need something to put though the pore. Most plastics are lighter than water and will just float and wont want to go through the pore. PMMA (acrylic) isn’t. But where I am it’s difficult to find PMMA pellets/granules.

So, with a pair of old wire cutters I cut up some acrylic sheet:

Then we can build our fludic system. We’re essentially going to build what was shown in the schematic at the top of the page. But there’s something missing, a way to drive particles through the pore. In this setup, we’ll use gravity.

The bottle cap will sit on top of beaker (or jar). It needs to be held in place, so the bottom of the cap with the hole in is completely submerged. But most of the bottle cap needs to be above the water line. If we add solution to the bottle cap it will then flow down into the bottom beaker.

For my setup I added a pump (cheap peristaltic pump from ebay) to recirculate the solution. Here’s the setup in schematic form. This is essentially the same figure as shown at the top of the page:

Below is the actual setup in all it’s glory! In my setup I used 1M KCl, about 7.5g of KCl in 100mL of DI water. I used AgCl electrodes (silver wire that’s been sitting in bleach for ~30mins).

I also needed to add a small barrier to help direct particles to the hole. Without this particles tend to get trapped in the fluid flow and spin in circles around the pore without going down it. I used a small piece of microscope cover glass:

You can then put a bias voltage across the electrodes and monitor the current. I used a oscilloscope to view the current trace. The scope has a built in function generator which I used to generate a 1v DC bias voltage.

Current was monitored using a self built transimpedance amplifier with a 1K feedback resistor. This was based around an LMP7721. But probably any FET opamp in a transimpedance configuration will work.

In the video above you can see the system working. As I pipette in some solution, you should see a small black particle fall down into the beaker. At roughly the same time, you should see a small upward “blip” on the oscilloscope, as the particle translocates the pore.

The PMMA particles tend to float on the surface tension. So I’m pipetting solution on top of the particle to make it sink.

Hopefully this provides a basic visual demonstration of the “pore” or coulter counter sensing approach. If you’re interested in ionic sensing, you should signup for Demonpore. Demonpore is making a molecular gaming platform which lets users experiment with pore based sensing. And my desire to experiment with macropores here comes out of discussions we’ve had around explaining nanopore sensing in visual and compelling ways.

The LX-4000 is the stage controller used in the Illumina Genome Analyzer 2. It’s a custom stage controller made by ASI somewhat similar to their ASITiger controllers. The controller is essentially a chassis with a backplane providing power and common bus to expansion cards. This post contains my notes on using this controller.

The LX-4000 contains 3 cards. The FW-1000 (filter wheel), a “Z/F” card (Objective, Z-axis), and a “X/Y” card. The “Z/F” and “X/Y” cards are essentially the same design. These PCBs are labelled “Two Axis Board” on my device one is “REV. C” (the X/Y) and the other is REV. D.

Communication

All 3 cards have ports that look like RS232 ports, only the port on the “Z/F” card can be used to communicate with the device. Commands for the other cards are routed over the internal backplane.

The FW-1000 card has a port labelled “RS-232” on my instrument this is not physically connected to anything. The filter wheel can be controlled manually by pressing the “NEXT” button or via the “Z/F” serial interface.

A CH340 based USB serial converter works well with the LX4000. Many CH340 RS232 converters work quite poorly, seemly mostly on the transmit side. A number of issues I had with getting the LX4000 to work with micro manager were caused by the device only registering some of the commands the CH340 was sending. I therefore recommend getting a better USB serial adapter. The Edgeport 8port serial card supplied with the Genome analyzer works well. Probably a (real) FTDI device would work well too.

The USB-serial adapter should be attached to the usually included “Reverser” (I believe a null modem converter). Baud rate is 115200 (8N1) on the “Z/F” card.

When powered up the serial will output “F” or “FX”. It may also report “MOTOR 1 NOT RESPONDING” or similar. This indicates that only one filter wheel is installed, and is not an error.

The general command set seems to follow the MS-2000 protocol.

While the instrument is initializing it’s possible to enter serial commands such as “1H I Z”, which will return information regarding the Z axis. However after a few seconds the device will start replying with “BELL”s (or possibly just in hex). I had issues getting the device to respond here.

However, the ASI TLX4000s.exe application will happily communicate with the controller and control everything except the X/Y stage without further modification (see notes on the X/Y stage below).

The LX4000 command set seems to be similar to the MS2000. However commands need to be prefix to address the required stage as follows:

1H: z stage
2H: x-y stage
3F: filter wheel

Micro manager is supposed to support the LX4000, but depending on the version you have you may have issues. I found that initial setup with the ASI software is most reliable. This software is a single exe called “TLX4000s.exe”, available from this excellent blog post on the GA2 ASI stages (or local copy). The software is somewhat troublesome to get working on modern systems and appears to have been written in Visual (Basic?) 6. The various runtime library either don’t work, or don’t include all the necessary components for Windows 10. My route to getting this working was to create a Windows 7 VM, then install Visual Studio 6. This doesn’t take long and insures that all the necessary libraries are installed.

In the genome analyzer the socket labeled “LIN ENC” on the X/Y card is as I understand it connected to the door sensor. The only pins connected on the genome analyzer are pins 5 and 8. This seems to provide a hardware lockout on the X/Y stage when the door is open. The following image shows a 1K resistor between these two pins, and when connected to the “LIN ENC” socket will allow the X/Y stage to operate. Take care when creating/installing this as shorting the wrong pins can damage the X/Y card (see “Cards” below).

Without this is place, the X/Y stage will not move on my unit, the Z stage and filter wheel will however happily work.

I was able to get the LX4000 to work with micromanager 1.4 and 2.0 (1.4.23.20160628 and MMSetup_64bit_2.0.0-gamma1_20210309). But things seem fairly unstable (much more stable with a good USB serial adapter, but auto-detection is still problematic). You need to add the following devices: XYStage/ASIStage ZStage/ASIStage and ASIFWController/ASIFW1000. Sometimes stages were detected correctly, sometimes not… (my config is here). I generally had to enter the COM parameters directly, scanning was not reliable.

Cards

It’s wise to be careful when shorting pins on the jumper connector shown above. Shorting the wrong pins dumps (likely RS232 level signals?) possibly on to the 5v power rail, or elsewhere. It’s possible to kill the card by doing this.

If this happens, it will likely blow a 5v trace near the expansion card connector. This traces are very thin and act like fuses… I’ve done this… repairing the trace seems to fix the card…

All the expansion cards seem to use Silicon Labs MCUs. C8051F122s on the two axis cards. C8051F015 on the FW1000.

The two axis cards uses TDA7269As to drive the stages.

Card communicate across the internal bus. The Z/F is the controller. If the Z/F card is not present the FW-1000 and X/Y cards will not come online (the LEDs and displays will not illuminate).

Jumpers SV3 and SV1 on the rear of the Z/F card control how control signals are routed across the backplane. If they are not present, the other cards will not come online. Here’s the default jumper configuration:

Here’s the default jumper configuration for the XY card:

There’s also a set of jumpers near the serial port. If these are set incorrectly you will experience varying degrees of weirdness (seems to still work in some configs, but outputs a lot of extra stuff). Here’s my config for the Z/F card:

And for the X/Y card:

Filter Wheel Config

The filter wheel can be configured with the TLX4000s.exe application (see Communication above). This can be used to change the filter wheel offset if it comes out of alignment. To do this, open the software, select “Filter Wheel” (after establishing communication). Then change the slider at the bottom of the “Filter Wheel Command” window. When you have adjusted the offset click on “RS” (Ram Save) which save settings so they are stored across reboot. The offset can likely also be adjusted using the “OF n” command where n is an offset, though I’ve not tried it.

The filter wheel has 8 positions (two of which are blanked out). The “Next” button on the LX4000 will only select between the two populated filtered. But serial commands can be used to move between any filter position. Micromanager supports this (use the ASIWheel device, and create groups to control filter wheel presets).

Power

The backplane provides +/- 15v, 5v and 24v to all cards. A 6 pin connector provides the +/- 15v and 5v supplies to the backplane. A second 3 pin connector supplies 24v.

In the image below the probe is on the 15v pin and from left to right voltages were probed as:

15v 5v 5v 0v 0v -15v

The connector to the right of the 6pin power connector is the 24v connector, on the opposite side of the board. This is supplied by a separate power supply that sits at the rear of the device (Condor GPFC110-24). In the orientation shown in the image below this reads as 24v 0v 0v.

Power is distributed to the expansion cards via the 8 thick traces shown in the image below on the right side of the expansion card. The last 8 traces probed as:

15V 5v 0.5v 0v 24v 0.2v -15v 0v

With the voltages <1v probably being measurement artifacts due to poor probing/grounding.

Of the 3 pins holes across the expansion connector, the central pin is not fitted. The pins on the edge (“top” and “bottom”) of the connector as connected. This is true for both the power and data lines.

Why was Manteia Important?

There are three main components to the sequencing approach used by Illumina, originally developed by Solexa in 2004:

• Sequencing-by-synthesis (detection of nucleotides incorporated by a polymerase)
• Reversibly terminated, labelled nucleotides
• Cluster generation through bridge amplification.

The general concept of sequencing-by-synthesis had been around since at least 1989.

Reversible terminators, in particular the 3′-O-Azidomethyl terminators used by Solexa were originally reported in 1991. Solexa modified these to incorporate a cleavable fluorescent label.

That leaves one final missing piece, cluster generation. Without an amplification approach Solexa would have had to use single molecule detection. Single molecule detection was significantly harder in the late 1990s. And even now, single molecule sequencing shows error rates significantly higher than achievable with clonal approaches.

Solexa acquired this technology from Swiss startup Manteia Predictive Medicine for, at an estimate, 1.5M USD [1]. This is a shockingly low price for a technology now foundational to a billion dollar market… but reports suggest that Manteia owner Serono had lost strategic interest in sequencing, which may explain the low acquisition price.

What was Manteia?

Manteia was a spin out of Serono, founded by Dr Pascal Mayer. They appear to have been pretty forward thinking, and were part of a small crop of next-gen sequencing startups that appeared in the late 1990s (there are now over 40).

Most of the technical information here, comes from a couple of presentations on slideshare [2] [3]. But there’s also ample information in their patent filings.

The basic approach suggested by Manteia, is pretty familiar, amplify single molecules on a solid surface, and sequence.

And indeed, the bridge amplification approach described, closely matches that presented by Illumina today:

The image also very look very much like those from early genome analyzer 1 and 2 instruments. In fact look pretty similar to today’s Miseq images. In fact, Manteia images look slightly cleaner. Reflect the fact they they used a single channel chemistry, showing no crosstalk.

The Manteia sequencing approach was far different than that used by Solexa however. While Manteia were also proposing sequencing-by-synthesis, this presentation doesn’t show fluorophores being cleaved. In fact, it look like the dyes stay attached, and signal is ever increasing.

Manteia would therefore have had to detect stepwise increases in signal. The chemistry appears to lack reversible terminators. Homopolyers would therefore cause larger jumps in signal intensity. Precise determination of homopolymer length would probably have been difficult, a problem that still exists in some approaches (Ion Torrent).

As for Illumina sequencing, phasing (clusters getting out of sync as they fail to incorporate) would be an issue. But the lack of reversible terminators makes phasing errors harder to compensate for algorithmically.

Photobleaching, would effect not only the current base, but signals for all prior incorporations. This would need to be compensated for and could be another source of error.

Ultimately, an ever increasing “background” intensity would also limit read length.

Overall, these issues seem challenging. Switching this chemistry over to a reversible terminator/cleavable dye approach was likely key to Solexa’s development of a viable sequencing platform.

But Manteia do appear to have generated proof of concept sequencing results:

Unfortunately, we’ll never know how Manteia might have progressed the platform based on these early sequencing results. As development at Manteia stopped before the platform was fully realized.

What does this means for Illumina today?

Manteia initially reported cluster generation in 1998. The earliest IP I could find referring to the bridge amplification approach was from September 1998, and expired in 2020.

Illumina is currently using Solexa’s reversible terminator IP to block the sale of MGI instruments in the US. That IP appears to expire in 2024.

After this point, it seems likely that other players will be able to act on this IP and build essentially, Solexa-style SBS sequencers.

Current Illumina instruments have progressed somewhat. The 2-color sequencing, and exclusion amplification IP is still active as far as I’m aware.

However even without this IP, it’s likely that a respectable Illumina-clone could be assembled based on the original Solexa approach.

It therefore seems likely that new “Solexa-style” approaches will appear in the not too distant future. And at the very least, MGI will be able to sell instruments in the US again.

This broadening of the sequencing market is likely what’s driving Illumina push into diagnostic applications. Recent acquisitions of GRAIL, and Verinata Health give them a route into clinical diagnostics. Where they not only own the core sequencing technology, but the diagnostic application too. This allows them to take a larger share of the profit, and more closely lock-in a diagnostic customer base.

Notes/References

[1] Solexa’s 2004 accounts note that “These rights were purchased from Manteia SA under terms of a joint asset purchase agreement with Lynx Therapeutics”. Overall it looks likes the Manteia rights cost them somewhere in the region of 1.5M USD, that is unless payments from the recently merged Lynx side of Solexa hide additional payments.

[2] “A very large scale, high throughput and low cost DNA sequencing method based on a new 2-dimensional DNA auto-patterning process”,P. Mayer, L. Farinelli, G. Matton, C. Adessi, G. Turcatti, J.J. Mermod, E. Kawashima, presented at the Fith International Automation in Mapping and DNA Sequencing Conference, St. Louis (MI,USA), October 7-10, 1998. Invited presentation (P. Mayer). Colonies from the 1998 presentation:

[3] A non confidential corporate presentation of “Manteia Predictive Médicine” as of September 2003. Présents DNA colony sequencing resutls, instrument, DNA preparation for genotyping.

It’s been a while since we heard much from Direct Genomics. This Shenzhen based sequencing company was a reboot of early NGS player and Quake spinout Helicos.

The company is also notable for being founded by Jiankui He. He is the scientist responsible for the germ-line genetic editing of 3 human babies. And is reportedly serving a 3 year jail sentence.

One of the last papers referring to the Direct Genomics sequencing platform is this one from 2017. Where they discuss the Direct Genomics GenoCare single molecule sequencing platform.

With He’s jail sentence the future of Direct Genomics seemed uncertain. But it looks like the GenoCare platform still exists. Now under development by GeneMind, a Shenzhen based Biotech company founded in 2012. A couple of papers appearing this year describing the platform under their stewardship.

In particular a medRxiv paper from September this year, describes a new two color single molecule sequencing approach.

This seemingly uses the same virtual terminators as the original Helicos approach, with the dye and “terminator” sitting on the base:

GenoCare two color sequencing isn’t the same as the Illumina two color approach, and doesn’t appear to provide the same advantages. On GenoCare’s platform two terminators (C and T) are labelled with a green dye, and two others (G and A) with a red dye [1].

This means that of each cycle, 4 images still need to be acquired. One set after flowing in C and A. And another after cleaving the dyes and flowing in T and G [2]:

In contrast to this, Illumina two color sequencing incorporates information from “dark” bases. This means they can take only two images, and perform no intermediate chemistry in their two color approach:

The GenoCare two color approach therefore doesn’t provide any advantage in terms of imaging speed/performance. Though the optics is likely cheaper than a four color system, it’s unclear as to why they don’t use a two color approach similar to Illumina’s which would allow them to take only two images, with no intermediate cleavage step on the same optical system. Perhaps there are IP issues here.

The GenoCare optical system uses objective style TIRF and a sCMOS camera. This appears to generate reasonable quality images, showing an SNR of >10. There have been huge advances in single molecule imaging since Helicos and this should simplify imaging/data analysis.

However, this doesn’t seem to have translated into massive improvements in data quality. With the GenoCare system showing mismatch, insertion, and deletion error rates of 0.61%, 1.45%, and 2.76% respectively.

Reads are generally short, but usable. With a lopsided distribution heavy in short reads:

Overall, the GenoCare system has an error rate one to two orders of magnitude higher than the market dominating Illumina approach. Their optical system is likely more expensive (requiring single molecule sensitivity). And read lengths are shorter.

The sole advantage of this platform is that amplification is not required. This may simplify sample preparation, and result in lower bias for certain applications. Of course also the case for other single molecule approaches (PacBio, nanopore).

Overall, it’s difficult to see how this platform could be competitive with Illumina (or the largely similar SBS approach used by MGI)… but perhaps they’ll manage to find a way forward.

Notes

[1] “Two terminators are labeled with a green dye, whose peak fluorescent emission wavelength is 552 nm, while the other two terminators are labeled with a red dye with peak fluorescent emission at 664 nm.”. The paper doesn’t appear to not which dyes are associated with which bases, but this combination seems likely from the figures.

[2] The exact combination isn’t clear from the paper, but they can only image one red and one green labelled terminator per imaging cycle to generate an unambiguous read.