Detection of miRNA and Proteins Using Silicon Nanowire Biosensors
Sep 5, 2023
"Detection of miRNA and Proteins Using Silicon Nanowire Biosensors"
Brian Dorvel, Biophysics
CNST Nanotechnology Workshop 2013
All right, so yeah, I'm going to basically talk about, as Dr. Bashir said, detection
of micro RNA and proteins using silicon nanowire biosensors.
So I just wanted to go through some slides real briefly about different cancer biomarkers
and what we're trying to look for, and what other universities and clinical settings are
also trying to look for.
And so it's a typical environment.
You have the host, the tumor microenvironment, and cancerous cells.
And there's basically two main ways that people are trying to understand cancer, understand
the biomarkers, and be able to diagnose and create treatments.
And one of them comes through high-throughput sequencing, such as using Illumina's system
or the other electronic detection methods in development right now, like IonTorrent,
to look at DNA and RNA sequences and look for the genes and the proteins expressed from
these.
And the other part is using biosensor-based platforms or point-of-care systems, which
look for not only RNA but other altered levels of protein expressions, different outside
markers for immune cell profiles, and also metabolites from cancerous cells.
And so just talking about the point-of-care diagnostics and why people are starting to
look into these a lot more is that there's a lot less, should I say, a lot more decentralization
of testing.
So you can get out of the core testing facility and monitor your treatment at home or in a
doctor's office a lot easier, and reduce the time of result and reduce the labor costs.
And second is increasing the sensitivity.
If you're able to increase sensitivity and have a smaller device as well, you might be
able to catch tumors in what they call malignancy or metastasis in an infant stage, where it's
going to be much easier to treat.
So this is just one of the DNA sequencers that's based upon electronic detection, which
is the main basis for my talk, is using electronic methods for doing protein-based detection
and RNA and DNA detection.
So this is Ion Torrent's electronic sequencing biochip, just as an example.
And basically this is what it looks like right here, where they have a bead and they attach
DNA to it.
And each time that a base is added through polymerase, it generates a response on this
metal oxide sensitive layer here.
And you can just see that the difference is this is essentially the area where this
bead goes, and this is what a M1 to M8 process for an Intel 32 nanometer node flow looks like.
You can see the incredible similarities.
And so right now they're able to get maybe 1.5 million sensors per chip on here.
And this is what they're actually trying to market now also to the clinical setting.
And that's what one of the chips looks like.
And so this is what the end goal is, to be able to detect not only proteins and RNA through
this method, but not only through pH, but also microRNA and proteins through their intrinsic
charge.
And this is basically our basic platform is ISFETs, or Ion Sensitive Field Effect Transistors
or biosensors.
And so basically the equations for a typical MOSFET and an ISFET are essentially the same
except for this metal is now replaced with a reference electrode and a salt solution.
And so what happens is you have an active area made functional by mobilizing probes
or antibodies, and the target molecules have a net change.
And what will happen is when a charged molecule binds to the surface, it alters the charge
density and the effective surface potential, and you get a shift in the device's response.
And so you get a change in the threshold voltage and an electronic-based detection event.
And so right now people have been trying to use CMOS-compatible nanowires or very small-width
devices because they're about the same size as probe molecules and targets, so you get
a better response than off of a planar device itself.
And so they've been able to use CMOS-compatible nanowires here for detection of antibodies,
cancer biomarkers, DNA, and even monitoring kinetics.
And so, but what is needed to create this to an actual marketable end is understanding
better the nanowire operation, including the fabrication and stability, and also its surface
functionalization response.
And also because for fluorescent methods, such as ELISA's, there's been very standard
protocols for immunohistochemistry that have developed, but yet with nanowires it's actually
in, you know, an infant stage.
And so a lot of protocols do need to be developed that people can work off of and standardize.
And so one of the main things of my talk is doing the copper characterization of the nanowires
and being able to, you know, digress that knowledge to other groups as well to be able
to improve the enhancement and the sensitivity of the devices.
And so one of the main things we came across was functionalizing the surfaces.
Now a lot of molecules, like APTIS here, contain three functional groups, which lead to polymeric
layers.
And when you're dealing with an electronic device that's interface sensitive and you
want to get as close as possible to the interface, this is detrimental to your sensitivity.
And so what we did is we came up with a vapor-based deposition method that uses a silane, but
it only has one functional linker here.
So it can either only form a diamond with itself or react with the surface, which we
were able to achieve actual true uniform high-density layers that were only one monolayer thick.
And just conjugation to these devices using either an epoxy moiety or an amino moiety,
you can use NHS esters, isothiocyanates, or amine groups.
And we all get very high-density levels.
And because it's only 15 minutes, we actually characterized this in detail and found that
you do get almost a 95% coverage due to looking at XPS densities of atomic elements.
And so this actually is being used in other current technology in other groups right now
to make reference sensors.
So they block out all these groups instead with a chain that's pH insensitive.
And they're actually able to get less than a millivolt per pH response when they deposit
this as opposed to the maximum sensitivity.
So it does a very good job of actually blocking these sites.
And so the other part is that we're trying to deal with the bio-interfacing and the fabrication
dependence on the sensitivity.
And so this is just what some of our nanowires here with 100 nanometers and some of the nanoplates.
And so these also have an ultra-thin thickness of maybe 30 nanometers, which allows it to
be a fully depleted device and give us the best sensitivity.
And so our study was to generate protocols for creating stable nanowires and functionalizing
the surface for protein-based sensing.
And also understanding the effect of the different variables like thickness and linker chemistry.
And so we took that vapor-based method and we tried to functionalize the surface and
optimize it with different functional groups here, such as this small linker here.
This one that has a peg.
And then typical glutaraldehyde, because usually the farther away you get it from the interface
with an electronic-based device, the lower your sensitivity is.
But what we found is that when we did the fluorescent measurements on this, we find
that this smaller linker actually provides the best and the highest density, but gives
us less signal-to-noise ratio than this pegylated one, which has, I guess, a better passivation.
And so this was the first step into optimizing and getting a better device for sensing is
using this chemistry.
And so next we just looked into primary antibody concentrations and using a standard ELISA-based
protocol for depositing primary antibodies.
So we found that if we go to 100 nanograms per mL, this is a change in that threshold
voltage I was talking about plotted versus that concentration.
And so we can see that when you go over 100 nanograms per mL, you essentially saturate
out the surface.
And you can see that with these different linkers, that the DSC one, which gave the
highest fluorescence, also gave the highest change in the threshold voltage.
And this also follows through the fluorescent images of the nanowire areas here.
And so we can just see that nonspecific binding with BS-PEG.
With all of these different, what is, rabbit IgG, mouse IgG, mouse IgA, what we get is
we get specificities over 20 to 1 versus similar IgGs and isotypes of immunoglobulins with the
BS-PEG-based chemistry.
And so here we can see just by looking at the gate oxide thickness of the device from
50 angstroms to 90 to 150, we find that looking at the current versus voltage curves that
we get, you know, as expected with a thicker oxide thickness, a worse sub-threshold slope
going from about 290 to 390 millivolts per decade, which is something that we decide
to change in the next generation of devices.
But you can see here that we're able to begin limits of detection of actual mouse IgGs to
between 0.1 and 1 gigogram per mL at least, using a standard silicon oxide, thermal oxide
based process.
But some of the issues that we did come across was, even though we have a thick oxide, the
sub-threshold slope indeed would be poor and we would get a lot of drift due to mobile
charges and interface traps.
But then when you go to a thinner oxide, we would get very bad leakage.
And so basically the way that we wanted to get around this was by then using high-k dielectrics
because we can increase the thickness of the layer while maintaining a very high capacitance
still.
And so we decided to start using hafnium oxide as our high-k dielectric.
And we used a 15 nanometer thick HFO2 ALD atomic layer deposited with an equivalent
oxide thickness of about 3.2 nanometers, which we did certain anneals and understood the
annealing parameters with forming gases and others to optimize the process to get very
high reproducibility, low drift, and low hysteresis.
And so basically we used this AG-AGCL reference electrode and do the same thing as we did
before and monitored the threshold voltage.
And as you saw before, it was maybe 290 to 390 millivolts per decade with a thermal oxide.
But with a hafnium oxide based system, we get a sub-threshold slope of 112 millivolts
per decade with a change in the threshold voltage of maybe 2.7 millivolts.
So we have a very good device performance in fluid with much better parameters than
what you would have with a thermal oxide device alone.
And so we're able to maintain that low gate leakage while having a high sub-threshold
slope and a low drift with a hafnium oxide system.
And so with hafnium oxide, we actually functionalized with poly-L-lysine because we found that even
though the SiO2 amino-based vapor deposition, you couldn't reuse the devices as well than
if you used an actual electrostatically absorbed layer like this.
And so we characterized this poly-L-lysine layer that we put down with two different
molecular weights, 9 to 14K and 70 to 150,000 molecular weights and found that we got actual
thicknesses of about 1 nanometer for each one.
But you can see that the roughness and the uniformity is less on the higher molecular
weight system.
And so we used a lower molecular weight system here to, and these are actually AFM measurements
where we scratched the surface and then measured the thickness afterwards.
And so we put down probe molecules and blocked it accordingly with standard based microarray
protocols and so we looked at a matched target here of microRNA10B versus a mismatched target
of microRNA21 for a nanowire and those wider nanoplate devices.
And what we found is that we were able to get limits of detection down to 1 to 10 femtomolar
with that low molecular weight poly-L-lysine and you can see that all the way through almost
seven orders of magnitude we get minimal binding of the mismatched target.
And so our work in progress as Dr. Bashir talked about we are now actually partnering
with Taiwan Semiconductor to create extended gate FETs with dielectrics including hafnium
oxide and there is a titanium nitride layer here.
So right now we are currently testing a 0.3 micron CMOS process and also some other processes
in imaging sensors for biological applications.
And we are using lots of different width length ratios and trying these experiments right
now in real time with microRNA as well as proteins.
So I tried to get through all of those reasonably quickly and I thank you guys for coming and
any questions or comments?
Questions?
Oh right now at Dow I don't know yet but it's the same one as this at Gmail right now.
So Brian I have a question.
Sure.
So how is it?
Yeah so what we found was that using a more hydrophobic blocking agent we were able to
get better specificity than if we had something that was more hydrophilic.
And also I think using stuff that was more negatively charged we got better specificities
than if we used agents that were more positively charged.
And also with that pegylated linker right there I feel that the peg layer that was there
was why we got the best specificities compared to the others.
Is that just because of the lower amine density so we'll have less charging actions?
Yeah and also I think that when you put down that peg linker if any of it hydrolyzes you
have a peg layer which is inert.
And so the non-specific binding is much lower than other unreacted sites.
Any other questions?
Okay no questions.
Let's thank all the speakers one more time.

