Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
Phil. Trans. R. Soc. A 370, 2474–2488
doi:10.1098/rsta.2011.0503
ImmunoFET feasibility in physiological salt
environments
BY PATRICIA CASAL1,5 , XUEJIN WEN2 , SAMIT GUPTA1,5 , THEODORE
NICHOLSON III1,5 , YUJI WANG2 , ANDREW THEISS1,5 , BHARAT BHUSHAN3 ,
LEONARD BRILLSON2 , WU LU2,∗ AND STEPHEN C. LEE1,4,5, *
1 Department of Biomedical Engineering, 2 Department of Electrical and
Computer Engineering, 3 Department of Mechanical Engineering, 4 Department
of Chemical and Biomolecular Engineering, and 5 Davis Heart and Lung
Research Institute, The Ohio State University, Columbus, OH 43210, USA
Field-effect transistors (FETs) are solid-state electrical devices featuring current sources,
current drains and semiconductor channels through which charge carriers migrate. FETs
can be inexpensive, detect analyte without label, exhibit exponential responses to surface
potential changes mediated by analyte binding, require limited sample preparation and
operate in real time. ImmunoFETs for protein sensing deploy bioaffinity elements on
their channels (antibodies), analyte binding to which modulates immunoFET electrical
properties. Historically, immunoFETs were assessed infeasible owing to ion shielding in
physiological environments. We demonstrate reliable immunoFET sensing of chemokines
by relatively ion-impermeable III-nitride immunoHFETs (heterojunction FETs) in
physiological buffers. Data show that the specificity of detection follows the specificity of
the antibodies used as receptors, allowing us to discriminate between individual highly
related protein species (human and murine CXCL9) as well as mixed samples of analytes
(native and biotinylated CXCL9). These capabilities demonstrate that immunoHFETs
can be feasible, contrary to classical FET-sensing assessment. FET protein sensors may
lead to point-of-care diagnostics that are faster and cheaper than immunoassay in clinical,
biotechnological and environmental applications.
Keywords: biosensing; immunoFET; protein detection; immunoassay; AlGaN; CXCL9
1. Introduction
Biosensing modalities can be of pivotal utility in clinical settings. Biosensors
detecting appropriate analytes, with sensitivities and modes of operation, can
potentially detect incipient disease prior to manifestation of clinical symptoms,
*Authors for correspondence ([email protected]; [email protected]).
Electronic supplementary material is available at http://dx.doi.org/10.1098/rsta.2011.0503 or via
http://rsta.royalsocietypublishing.org.
One contribution of 10 to a Theme Issue ‘Biosensors: surface structures and materials’.
2474
This journal is © 2012 The Royal Society
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
ImmunoFETs in physiological buffer
2475
reducing patient morbidity and mitigating mortality. Optimally, such sensors
should operate non- or minimally invasively, such that the act of sensing itself
does not impose significant morbidity on the patient.
Biosensors couple a bioelement (receptor that provides recognition of a
specific analyte) with a transducer element to convert a biological event (e.g.
binding of analyte to receptor, production of an enzymatic reaction product,
consumption of an enzyme substrate, etc.) into an electrically measurable signal
[1]. This general paradigm has been manifest in electrochemical biosensors
since the earliest attempts to sense biologically important small molecules or
macromolecules, and was followed in the first electrochemical biosensor (the
glucose enzyme electrode [2]). The glucose enzyme electrode was followed
relatively quickly (in 1970 [3]) by the development of the earliest ion selective
field-effect transistors (ISFETs). ISFET design was significantly refined within
a decade of inception of the sensing modality [4,5]. ISFETs feature integration
of bioaffinity/catalytic elements (in ISFETs, typically an enzyme) deployed in
the place of a gate on the capacitance layer of a metal oxide semiconductor
field-effect transistor (MOSFET). Basic ISFET operation is broadly similar to
operation of conventional FETs, excepting that ISFET current modulation is
provided by ions produced by the enzyme and not by an electrical bias on a
gate electrode. This coupling of biological affinity elements (receptors, antibodies,
other proteinaceous affinity elements, nucleic acids of various descriptions, etc.)
with the field-effect modulation principle of semiconductor devices was enticing,
offering the promise of rapid, highly sensitive detection using a sensing platform
(i.e. the MOSFET) that had been highly developed in the electronics industry.
The ability of ISFETs to detect ionic charges further suggested the possibility
that charged biological macromolecules (e.g. proteins, nucleic acids) might also
be detectable by FETs.
The measurement of proteins via their intrinsic charges was first proposed
by Schenk [6] when he envisioned deploying a receptor layer of antibodies
with specific affinity for an antigen of interest (i.e. protein analyte) on the
gate region of a FET: an immunoFET. One of the earliest immunoFET-like
sensors featured immobilized anti-human serum albumin (anti-HSA) IgG on a
polyvinylbutyral (PVB) membrane overlying an ISFET-sensing channel [7]. This
sensor was capable of detection of HSA in solution (pH 7.0, 37◦ C) with a 2 mV
shift in Ids (current, drain to source). Sensor signal was linear with analyte
concentration over a range of 0.01–1.0 mg ml−1 , with a lower detection limit of
0.01 mg ml−1 . While these experiments demonstrated that proteins could indeed
be detected via immunoFET, the devices were not satisfactorily sensitive for
many medical, biological and biotechnological applications. Limited sensitivity
was a key limitation of many early immunoFETs [8–12], which impeded their
translation to clinical devices.
The goal of optimizing immunoFET sensitivities to be comparable to those
offered by existing immunoassays (i.e. enzyme-linked immunosorbent assays
(ELISAs)) was elusive. Lack of a sensitive immunoFET led to promulgation
of a hypothetical, theoretical rationale for the sensitivity limitations of the
platform [13,14]. This model purported to define the limited sensitivity
as being unavoidable and, in fact, intrinsic to all immunoFETs. Bergveld
et al. [14] opined that ‘it should be concluded that a direct detection of
protein charge is impossible’ and, further, that the idea that a layer of
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2476
P. Casal et al.
(a)
S
e–
D
e–
D
(b)
S
Figure 1. (a) Antibodies (Y shapes) deployed on an immunoFET surface as depicted in classical
representation [13,15,16]. Antibodies attached via a polymeric linker (not shown) and an analyte
(circle) bound to antibody. Approximate Debye length position is indicated by the dashed line.
Uniform orientation of antibody molecules on an immunoFET surface assumed in classical analysis,
with antigen-binding domains uniformly pointing 180◦ away from the sensing surface. (b) A more
realistic depiction of antibody alignment on an immunoFET. Note the lack of consistent orientation
of antibodies relative to the FET surface.
charged proteins at a FET surface could modulate the electric field in
the semiconductor ‘should definitely be forgotten’ [15]. This is the classical
immunoFET assessment [13,15–18].
Classical assessment asserts that buffer counter-ions between the bound analyte
and sensing surface shield analyte charges to the extent that those bound
charges cannot influence current flowing through the FET semiconductor channel.
The counter-ions were hypothesized to result in the formation of an electric
double-layer between the analyte and the sensing surface having the thickness
of the Debye length (inversely proportional to ionic strength). In physiological
solutions (approx. 150 mM Na+ ), Debye length is short: of the order of a couple
of nanometres. Since some proteins (specifically, antibodies) have dimensional
aspects of the order of 10–12 nm, it was argued that charges of proteins bound
to antibody receptors would be held at distances greater than the Debye length
from the sensor surface, making electrostatic detection of such charges impossible.
Figure 1a shows a schematic of the classical view.
The classical model was readily accepted to rationalize immunoFET failures,
and has long held sway in discussions of immunoFET protein sensing.
Unfortunately, the classical model is marred by multiple misconceptions about
the properties of antibodies deployed on synthetic surfaces. Classical analysis
misrepresents not only antibody adsorption properties, but also antibody
structure and dynamics. These misconceptions render the classical model
incapable of accurately describing the actual behaviour of immunoFETs.
Classical analysis hinges on the distance between the extreme ends of antibody
variable domains and C3 domains (figure 2). The distance is about 10 nm,
substantially greater than the predicted Debye length under physiological salt
concentrations. Bergveld reasoned that charges of analytes bound to antibody
receptors would be shielded from the sensing channel of the FET by buffer
counter-ions, and therefore not be detectable.
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
ImmunoFETs in physiological buffer
VH
VL
2477
variable
region
CH1
CL
CH2
hinge
region
constant
region
CH3
Figure 2. Schematic of an IgG antibody. Heavy and light chains of the variable region are labelled
as VH and VL , respectively. The heavy chains of the constant region are labelled as CH 1, CH 2 and
CH 3, while the light chain is CL . The hinge region is indicated with an arrow. Analyte (or antigen)
binding interaction occurs at the variable domain.
The reasoning is immunologically unsound for multiple reasons, though we
focus primarily on considerations that are applicable to all immunoFET sensors.
For classical analysis to be valid, antibodies must (i) be rigid bodies that (ii)
adsorb to the sensing surface solely through their terminal C3 domain (figure 2).
Logically, failure of antibodies to conform to either of these conditions calls
classical analysis into question. In fact, antibodies adsorb to surfaces via a
nearly random distribution of their structural domains [19], depicted in figure 1b.
Moreover, antibodies are highly flexible, able to bend through arcs of 180◦ or
more, with additional flexibility provided by the so-called ‘molecular ball and
socket’ region occurring between the V and C1 domains [19]. These facts are
discussed in most immunology textbooks (for example [19]), and have been for
multiple years. These facts have implications for assessment of immunoFET
function.
The combination of antibody adsorption in variable orientations relative to
the surface and antibody flexibility causes bound analytes to be held in a
distribution of orientations and distances from the sensing surface. Some of
the analyte charges should thus be expected to be held within the Debye
length, and therefore analyte electrical fields should be detectable by FET. That
antibody adsorption to surfaces is not typically oriented, and that antibodies are
indeed highly flexible, is not only immunological orthodoxy but also provides a
rationale for successful immuno- and bioFETs [20–30]. These facts also indicate
that, when immunoFETs fail, the mechanism is probably not as described
by the classical model. This is important because understanding the mode
of failure can potentially drive remediation of the immunoFET design. These
misconceptions are pivotal in classical assessment, but their invalidity has been
widely appreciated in immunology for years [19].
Classical assessment further ignores potential interfacial design approaches to
minimize the distance between bound analyte charges and immunoFET sensing
surfaces to maximize sensitivity. Neither does classical assessment consider
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2478
P. Casal et al.
differentiable ion permeability of various FET platforms. Ion permeation alters
electrical properties of MOSFETs, impeding accurate sensing [31]. Less ionpermeable FETs (e.g. AlGaN/GaN HFETs) have long been available and were
used in the studies presented herein.
Beyond the above issues, the classical assessment depicts proteins as
point-charges. In reality, proteins are space-filling nanoscale objects with fixed
or highly constrained three-dimensional conformations whose charges are tightly
associated with individual amino acid (aa) residues. Specific distribution of charge
is determined by both analyte primary aa sequence and analyte conformation.
Furthermore, antibodies recognize specific epitopes of their cognate protein
antigens. Epitopes typically span regions of 10–12 aas, such that a protein of
100 aas is expected to have at least 10–12 distinct epitopes which could be bound
individually or simultaneously [19]. This fact may have interesting potential
implications for further development of immunoFET technology to detect charges
of proteins independent of one another for purposes of protein identification and
characterization (see §3).
While FET-based sensing is well accepted and widely used for the detection
of various macromolecules, its use for the detection of proteins is not. This is
due in part to the previous assertions that direct, label-free protein detection
was infeasible [13–15], and the frustrating history of immunoFET development.
This is unfortunate, in light of the fact that immunoFET assay detects an
analyte attribute directly (i.e. analyte charge or electrical field), without need
for secondary reagents (e.g. enzymes, chromogenic or radioactive substrates).
This could make immunoFET assay potentially rapid, convenient and desirable,
providing that immunoFETs operate with adequate sensitivity in physiological
salt environments. In contrast, optical and other detection modalities that are
available for proteins are often cumbersome, requiring multiple, time-consuming
steps followed by the addition of secondary fluorescent labels or radioisotopically
labelled compounds. In addition to providing real-time, direct detection, FET
platforms are made by standard semiconductor processing methods, and are
conveniently size-scalable.
ImmunoFET sensing has clear biomedical application in multiple disease states
(in the presented work, for transplant applications), and could be applied to
any biotechnology or environmental use where detection of a protein analyte
is required. For example, immunoFETs might be used to detect proteinaceous
environmental toxins and agents for biodefence applications (e.g. detection
of biological warfare substances or proteins modified by them). Accurate
consideration of the properties of the biological and polymeric components of
immunoFETs, as well as immunoFET sensing data, shows clearly that genuine
immunoFETs are not infeasible as a class [20–22]. We show here multiple
functional immunoFETs operating physiological buffers, contrary to predictions
of classical analysis.
Our analytes here are monokine induced by interferon-g (MIG, CXCL9) of
humans and mice. In both species, MIG is a pro-inflammatory chemokine that
is a chemoattractant for cytotoxic T-cells. MIG increases during inflammatory
responses, rising from a normal concentration of 40–100 pM [31], to 1–2
orders of magnitude higher during acute inflammation [32]. In transplant
biology, rising graft MIG concentrations precede allograft rejection and can be
used as an early indicator of imminent rejection [33,34]. Human and murine
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
ImmunoFETs in physiological buffer
2479
MIG (huMIG, muMIG) have about 80 per cent protein sequence identity,
similar isoelectic points and perform similar physiological functions, but are
immunologically differentiable. Antibodies (IgG) specific for huMIG and muMIG
[35,36] were used to build species-selective immunoHFETs. Results show that
receptor/analyte properties define immunoFET specificities in predictable ways,
a finding that is difficult to rationalize, unless planar immunoHFETs can
in fact function in physiological buffers. This report extends our previous
paper demonstrating feasibility of bioFETs [20] into the arena of immunoFET
detection by demonstrating selective and sensitive immunoFETs. Multiple
planar immunoHFETs detecting muMIG and huMIG [37] in physiological buffer
provide proof.
2. Material and methods
(a) Chemicals
The silane polymer used was triethoxysilane aldehyde (TEA; United Chemical
Technologies, Bristol, PA), and aminopropyl triethoxysilane (APTES; Gelest,
Inc., Morrisville, PA). Polyclonal anti-muMIG IgG was from R&D Systems,
Inc. (Minneapolis, MN). Polyclonal anti-huMIG IgG, biotinylated anti-huMIG
IgG and recombinant huMIG and muMIG are from Peprotech, Inc. (Rocky
Hill, NJ). EZ-Link Sulfo-NHS biotin was purchased from Pierce, Inc. (Rockford,
IL). Streptavidin (SA), SA conjugated to horseradish peroxidase (SA-HRP) and
Dulbecco’s phosphate-buffered saline (PBS) containing 150 mM NaCl, pH 7.4,
are from Invitrogen, Inc. (Carlsbad, CA). o-Phenylenediamine dihydrochloride
(OPD) tablets were purchased from Sigma-Aldrich, Co. (St. Louis, MO).
(b) Transistor fabrication
AlGaN/GaN HFETs were constructed as previously described [20,30,38].
AlGaN/GaN heterostructures were purchased from CREE, Inc. (Raleigh, NC)
and surface oxidized via ion-coupled plasma treatment (ICP) by oxygen plasma
[30]. One hundred micrometres of the AlGaN barrier was recessed in a
Cl-based ICP plasma so that the threshold voltage of the device was shifted
to the −0.5 to +0.5V range. The conducting channel of the HFETs varied from
50 to 100 mm in width and length. The device reservoir with an average height
of 10–20 mm allowed the conducting channel access by the samples. The chemical
gate formed on the oxide was functionalized with receptors (antibodies or SA)
for specific analyte binding.
(c) Surface preparation and device sample exposure
For the huMIG and muMIG receptor-specificity experiments, AlGaN HFETs
were surface functionalized with 5 per cent TEA in ethanol (following the APTES
deposition protocol [39]) and subsequently 1 mg ml−1 anti-huMIG or anti-muMIG
IgGs. The sample reservoir was exposed to 15 ml of 5 mg ml−1 (0.43, 0.41 mM,
respectively) huMIG or muMIG.
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2480
P. Casal et al.
For experiments comparing detection of mixed biotinylated and native
MIG samples (bMIG and huMIG, respectively), two AlGaN HFETs were
used. One device was functionalized with 5 per cent APTES in ethanol,
biotinylated using 1 mg ml−1 biotin at 37◦ C and subsequently treated with
5 mg ml−1 SA. This device was used for bMIG detection using SA as
the specific recognition element to bind biotin on bMIG. The second
device was functionalized as described above with 5 per cent TEA
in ethanol and anti-huMIG IgG and used to detect MIG (both MIG
and bMIG).
(d) Electrical measurements
Three-terminal (source, drain and gate) current–voltage characteristics of
AlGaN/GaN HFETs were measured using an Agilent 4156C semiconductor
parameter analyser at room temperature. The source/drain current is modulated
by the gate bias to the order of 1 mA mm−1 for detections, so that the device
is working in the subthreshold regime for best device performance [28,29].
A device source/drain characteristic was first measured with PBS only (baseline
measurement). The protein solution was then applied with a micro-pipette
and incubated for 5 min, after which second source/drain characteristics were
measured for comparison. The charges introduced by the binding of analyte to
surface receptors modulate the source/drain current.
(e) Immunosorbent assay
ELISAs were performed in 96-well Nunc Maxisorb ELISA plates to corroborate
electrical sensor data. For huMIG versus muMIG detection, wells were incubated
with 1 mg ml−1 anti-huMIG or anti-muMIG IgG for 1 h at 37◦ C; background
wells were incubated with PBS. Wells were then blocked with 5 per cent bovine
serum albumin (BSA) in PBS for 2 h at 37◦ C before exposure to 10 ng ml−1
huMIG or muMIG for 10 min at 37◦ C. Subsequently, wells were incubated with
1 mg ml−1 biotinylated anti-huMIG or anti-muMIG IgG for 1 h at 37◦ C. Plates
were then incubated with 1 mg ml−1 SA-HRP in PBS for 1 h at 37◦ C. OPD was
freshly prepared according to the manufacturer’s directions and added to each
well for 20 min; the reaction was stopped with 3M H2 SO4 , and the absorbance
of the reacted OPD solutions were measured in a Victor X3 Plate Reader
(spectrophotometer from Perkin-Elmer) at 490 nm. Wells were rinsed five times
in 0.1 per cent Tween-20 in PBS between each step.
For confirmation of bMIG/MIG electrical sensing data, ELISAs similar to those
described above were performed. In these tests, wells were treated as above with
anti-huMIG IgG for MIG detection. Treatment for these tests varied in that
the wells were exposed to solutions of bMIG/MIG in varying ratios as opposed
to solely huMIG. Wells were also treated for detection of bMIG using SA as the
detection agent. Wells were incubated with 500 ng ml−1 SA for 2 h at 37◦ C followed
by exposure to 1 ng ml−1 bMIG/MIG solutions (0–100% bMIG in 20% increments)
for 30 min at room temperature. Wells were then incubated in 1 mg ml−1 SA-HRP
and exposed to OPD as above. Absorbance of reacted OPD solutions for all wells
was measured in the Victor X3 Plate Reader at 490 nm.
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2481
ImmunoFETs in physiological buffer
(a)
(b)
absorbance
0.25
0.20
0.15
0.10
0.05
0
normalized current (Ids)
(c)
(d)
1.20
0.90
0.60
0.30
0
0.10
0.20
0.30
Vd
0.40
0.50
0
0.10
0.20
0.30
Vd
0.40
0.50
Figure 3. Detection of huMIG and muMIG by ELISA (a,b) and immunoHFET (c,d). ELISA
was used to demonstrate selectivity of antibodies to (a) huMIG and (b) muMIG. Reactivities
of (a) anti-huMIG antibodies with huMIG and (b) anti-muMIG antibodies with muMIG are
shown by grey bars. Reactivities of (a) anti-huMIG with muMIG and (b) anti-muMIG with
huMIG are shown by cross-hatched bars. Responses of immunoHFETs with (c) antibodies to
huMIG and (d) antibodies to muMIG decorating their respective sensing channels to PBS (solid
triangles), PBS with 5 mg ml−1 huMIG (open squares) and PBS with 5 mg ml−1 muMIG (open
circles). Current changes associated with species-matched MIG treatment reproduced within
10% (n = 3).
3. Results and discussion
huMIG and muMIG in PBS were quantified and the species selectivity of antihuMIG and anti-muMIG IgGs were corroborated by ELISA (figure 3). huMIG
and muMIG are positively charged (+19 charges/muMIG, +20 charges/huMIG,
pH 7. 4 [37]) and HFETs are n-type: their charge carriers are electrons. Therefore,
as expected, binding of huMIG or muMIG to sensing channels increased current
drain to source (Ids ; figure 3). However, the immunologically distinct MIG species
are detected differentially by immunoHFETs with corresponding species-specific
antibodies on their channels (figure 3).
Both murine and human-specific MIG immunoHFETs exhibited unchanged
Ids after exposure to PBS. Anti-huMIG immunoHFETs exposed to muMIG
gave responses similar to PBS background, and anti-muMIG immunoHFETs
exposed to huMIG also gave responses comparable to background. Conversely,
immunoHFETs decorated with anti-huMIG IgG exhibited approximately 22
per cent Ids increase on exposure to huMIG, and sensors with anti-muMIG
IgG exposed to muMIG exhibited approximately 15 per cent Ids increase
(figure 3). huMIG and muMIG minimal detection limits were similar (approx.
1 fM; electronic supplementary material, figure S1). Analyte detection specificities
of immunoHFETs reflect ELISA-demonstrated antibody-binding specificities
(figure 3), the though the assays are different (ELISA incorporates multiple
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2482
P. Casal et al.
secondary reagents that the immunoHFET assay does not). Having demonstrated
that homologous but immunologically distinct analytes are differentially detected
by immunoHFET, we next tested whether immunologically similar but distinct
analytes in single samples could be discriminated.
Native (unbiotinylated) and biotinylated huMIG were mixed at varying
stoichiometries and assayed by bioFET (SA as receptor), and immunoFET (antihuMIG IgG as receptor). Neither the bioFET nor the immunoFET responded to
PBS, nor did the SA bioFET respond to native MIG (electronic supplementary
material, figure S2). However, immunoHFETs with anti-huMIG on their sensing
channels detected huMIG regardless of biotinylation state (reflecting total
huMIG, bioinylated and native; figure 4), while streptavidin bioFETs detected
only biotinylated analyte, differentially detecting the varying concentrations of
biotinylated huMIG in samples (figure 4a,b).
Immuno/bioHFETs detect MIG at biologically meaningful concentrations
in physiological buffers and can be configured to detect analyte mixtures, or
single constituents of mixtures (figure 4). Results, though contrary to classical
assessment [13,15–18], are congruent with recent HFET protein-sensing results
[20,21,24,25,29,30,40]. Nonetheless, the notion that some immunoHFETs are
feasible in physiological environments is not widely accepted, motivating direct
demonstration of the proposition here.
Classical analysis [13,15–18] cannot model actual behaviour of immunoHFETs
because it misrepresents antibody properties. Both assumptions that antibodies
are rigid and uniformity of antibody orientation on immunoFET channels
(adhering to surfaces solely by C3 domains) are critical to classical immunoFET
analysis: negation of either assumption invalidates the central argument of
classical analysis. Immunologists have long known both assumptions are incorrect
[19], and our results reflect the invalidity of those assumptions.
IgG hinge region (between C2 and C1 domains) flexibility, known since
immunoglobulin structure was determined, allows individual arms of antibodies
(consisting of C1 and variable domains) to bend through arcs of up to 180◦
[19]. This conformational freedom is critical to antibody biology. Hinge flexibility
allows binding of multi-valent antigens, even if relative positions/orientations of
individual epitopes of antigens are irregular, facilitating formation of aggregates
for phagocytic uptake and elimination [41]. In immunoFETs using intact IgGs
as receptors, hinge flexibility should allow positioning of bound analytes in a
distribution of orientations/proximities relative to the rest of the antibody and
to the immunoFET sensing channel. This should occur whether antibodies are
consistently oriented on sensing channel surfaces or not.
Individual antibodies bind specific single epitopes of protein antigens, typically
10–12 aa long, often contiguous in antigen sequence. Through the use of antibody
fragments, to remove flexible hinge regions and reduce overall size, it may be
possible to preferentially position specific charged regions of analytes proximal to
sensing surfaces. This may allow for the detection of specific charged regions
of analytes as opposed to detection of analyte net charge. Use of epitopespecific orientation would lend itself to the detection of net neutral proteins via
exploitation of more highly charged regions.
Also, in the absence of specific affinity elements or chemoselective conjugation,
surface adsorption of antibodies is not consistently oriented. No biochemical
process forces antibodies to adsorb exclusively via the C3 (or any other) antibody
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2483
ImmunoFETs in physiological buffer
(a)
0.9
0.8
0.7
absorbance
0.6
0.5
0.4
0.3
0.2
0.1
0
0
20
40
60
%bMIG in solution
80
100
% change in current
(b) 160
120
80
40
0
20
40
60
80
100
120
%bMIG in solution
Figure 4. Differential detection of native and biotinylated huMIG in mixed samples by (a)
immunoassay and (b) immuno- and bioFET assays. (a) Primary analyte receptor (anti-huMIG
or SA, respectively) was deployed on ELISA plates followed by the secondary affinity element
(biotinylated anti-huMIG or SA-HRP, respectively). Hatched bars indicate biotinylated huMIG
content of samples; black bars represent total huMIG content of samples. (b) Per cent change
in current is from baseline to the exposed analyte source/drain characteristic. Note that part (a)
compares the absolute magnitude of absorbance, while part (b) compares the trend of device signal.
(a) Bars with horizontal lines, bMIG; filled bars, MIG. (b) Squares with continuous line, bMIG;
diamonds with dashed line, MIG.
domain: consistent alignment on surfaces requires modification of antibodies or
surfaces [42–44]. As for most immunoFETs, alignment of antibodies relative to
sensing channels was not attempted here, but may have potentially interesting
consequences.
Since adsorption does not occur exclusively at any specific antibody domain,
the calculated distance between the sensing surface and bound charges,
determined assuming uniform antibody adsorption to the surface by antibody C3
domains, and comparison of that distance with the predicted Debye length (the
distance over which counter-ion shielding should occur) in physiological buffer
cannot be relevant to immunoFET feasibility. However, the comparison (bound
analyte charges to sensing surface distance to Debye length) is the crux of the
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2484
P. Casal et al.
classical infeasibility argument. Hence, the infeasibility argument as originally
formulated is not germane to behaviour of immunoHFETs made as ours are.
That said, interfacial film structure is a primary determinant of sensitivity [39],
though not in the manner classical immunoFET assessment suggests.
Film morphology determines analyte charge to sensing channel distance, and,
thus, immunoFET sensitivity [39,45,46]. Consistent with theory [47], previous
data [20,29,30,39,40,48] show that immuno- and bioFET signal magnitude is
highly dependent on analyte charge to sensing channel proximity. Isrealachvili [47]
states that FET signal magnitude varies with the sixth power of charge-to-surface
distance. We cannot determine the mean distance of analyte charge to HFET
surface sufficiently accurately to empirically validate Isrealachvili’s prediction
with our immunoFETs, though it is clear that nanometre-scale changes in film
thickness, and in position of bound analytes in interfaces, profoundly influence
signal magnitude [29,30,39,40,48].
The sensors here were not engineered to minimize analyte-to-channel distance
and maximize sensitivity, but this critical parameter can be addressed by multiple
means. First, the sensor interfaces presented in this work use a trivalent silane
(TEA) that is similar to APTES. Therefore, we expect the interfacial height of
the TEA layer to be higher than the minimal height achieved using a monovalent
silane derivative (i.e. as with aminopropyl dimethylethoxy silane (APDMES)
[49]). In comparison with films made with trivalent silanes, monovalent silanes
form thinner polymer films with more regular (smoother) surfaces, shown in
figure 5 [49]. Based on differential sensitivity observed using APTES (trivalent)
and APDMES (monovalent) interfacial polymers [26,48], we expect that the
sensitivity of the presented devices will be enhanced by depositing more ideal
(i.e. thinner, more regular) interfaces built using monovalent silanes. That
said, the sensitivity of the unoptimized immunoHFETs demonstrated here was
sufficient to allow us to demonstrate immunoHFET feasibility in a physiological
environment.
Shielding of charges of bound analytes by buffer ions indeed occurs in
immunoFETs [20,29,30,39,40,48], but the key issue for immunoHFET feasibility
is whether charges of any given analyte bound to antibody receptors of an
immunoHFET are shielded by ions in physiological buffers beyond detection
by the underlying FET. This is a complex consideration (encompassing
analyte charge density, charge distribution, specific receptor and analyte threedimensional structures, specific receptor epitope recognition properties, receptor
bioconjugation conditions, interfacial film morphology, sensor dielectric thickness,
specifics of sensor operation, etc.), dependent on particulars of the immunoor bioFET at hand. We demonstrate multiple affinity elements (two antibodies
and SA) used to make functional HFET protein sensors and there is no reason
to believe that the affinity elements used here are special per se, allowing
configuration of multiple immuno/bioHFETs for diverse sensing tasks in diverse
environments.
Orientation of affinity elements might drive differences in HFET performance
should differential orientation influence positions of analyte charges relative to
the sensing channel. Interfaces with regular affinity element orientation can be
constructed [42–44]. Assuming the validity of theoretical exponential chargeto-surface distance relationships [47], it may be possible to use oriented, rigid
affinity elements to build immunoFETs that detect specific analyte charges
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2485
ImmunoFETs in physiological buffer
(a) 5
(b) 2.0
20
3
2
1
0
roughness (nm)
thickness (nm)
4
1.5
15
1.0
10
0.5
5
0
RMS
P-V
roughness (nm)
air
0
Figure 5. Interfacial silane films exhibit differential thickness and roughness as a function of their
siloxane valency. Silane deposition protocols for APTES and APDMES were as for TEA described
in §2. (a) Film thickness of trivalent (APTES) and monovalent (APDMES) silane polymer films
are compared, including SiO2 substrate as a reference. (b) Surface roughness estimated using RMS
(root mean square) and P-V (peak to valley) values after silane deposition on SiO2 wafers. Error
bars represent ±1 s.d. Adapted from Bhushan et al. [49]. Open bars, SiO2 (reference); bars with
horizontal lines, APTES; checked bars, APDMES.
or regions of charge in preference to or exclusion of others. ImmunoHFETs
using intact antibodies, and perhaps even some antibody fragments, as receptor
may not allow for detection of specific charged regions of analytes owing to
antibody (IgG) conformational freedom [19,41]. Intact IgGs would be expected
to hold analytes in a distribution of orientations and distances relative to the
surface, even if antibody binding to the FET surface was exclusively via the IgG
antibody C3 domain, as was originally, but inaccurately, assumed in classical
immunoFET analysis.
huMIG and muMIG are highly charged: analytes with lower net charges may
be less sensitively detected by immunoFET. It is unknown whether immunoFETs
detect net molecular charges or, as seems more likely, individual charges within
a critical distance of the sensor regardless of charge affiliation with individual
analyte molecules. The relationship of protein size, charge distribution and
conformation to detection by immunoFETs is also largely uninvestigated. We
have begun to manipulate the charge, multimerization state and size of huMIG
and other analytes by recombinant DNA and bioconjugate means to differentiate
these possibilities.
Differing FET architectures provide differing device electrical stabilities in
high salt environments [20], potentially affecting their amenability to in vivo
use. FET architecture also influences other immunoFET/bioFET properties,
as improvements in sensitivity from incorporation of a control gate show
[29,30,39,40,48]. AlGaN/GaN HFETs may be particularly suited to use in high
osmolarity environments (as in vivo) because of limited AlGaN permeability to
buffer ions and high-electron current drive properties [20]. Other FETs with, or
engineered to have, similar properties may be as efficacious and more economical
[39]. ImmunoFET economy and ease of fabrication may be important, as potential
for immunoFET sensors in clinical applications is large.
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2486
P. Casal et al.
4. Conclusions
The history of immunoFETs is illustrative of potential negative consequences
of dogmatism to technological progress. The classical assessment is flawed
in such a way that it can provide neither explanation nor rationale for
limitations of immunoFETs, nor, more importantly, guide solutions to technical
challenges for development of functioning immunoFETs. We provide proof for
immunoHFETs operating in physiological solution, using an ion-impermeable
platform (AlGaN/GaN HFETs) and show signal magnitude for a fixed analyte
charge to be directly related to charge proximity to the sensing channel
[20–22,29,30,39,40,44,48]. We believe recognition of the significance of these
nuances to immunoFET function will facilitate development of a myriad of
immunoFETs directed to other analytes working in high salt environments.
Given economical FET platforms, immunoFET assay could potentially supplant
more laborious, time-consuming and expensive immunoassays in clinical and
laboratory settings.
This work was made possible by a National Science Foundation grant (CBET 0758549) and a
Materials Research Institute Grant for Research from The Institute of Materials Research of The
Ohio State University. Additionally, we would like to thank the Ohio State University Medical
Center artist, Dennis Mathias, for his work in developing figure 1. The authors are grateful for
the critical role of their late friend and mentor Dr Charley G. Orosz of The Ohio State University
Division of Transplant for motivating them to develop sensors for MIG.
References
1 Thevenot, D., Toth, K., Durst, R. & Wilson, G. 2001 Electrochemical sensing: recommended
definitions and classifications. Biosens. Bioelectron. 16, 121–131. (doi:10.1016/S0956-5663
(01)00115-4)
2 Clark, L. & Lyons, C. 1962 Electrode systems for continuous monitoring in cardiovascular
surgery. Ann. N. Y. Acad. Sci. 102, 29–45. (doi:10.1111/j.1749-6632.1962.tb13623.x)
3 Bergveld, P. 1970 Development of an ion-sensitive solid-state device for neurophysiological
measurements. IEEE Trans. Biomed. Eng. 17, 70–1. (doi:10.1109/TBME.1970.4502688)
4 Matsuo, T. 1981 Methods of ISFET fabrication. Sens. Actuators 1, 77–96. (doi:10.1016/02506874(81)80006-6)
5 Matsuo, T. & Wise, K. 1974 An integrated field-effect electrode for biopotential recording.
IEEE Trans. Biomed. Eng. 21, 485–487. (doi:10.1109/TBME.1974.324338)
6 Schenck, J. 1980 Inventor field effect transistor for detection of biological reactions. US Patent
no. 4,238,757.
7 Gotoh, M., Suzuki, M., Kubo, I., Tamiya, E. & Karube, I. 1989 Immuno-FET sensor. J. Mol.
Catal. 53, 285–292. (doi:10.1016/0304-5102(89)80063-X)
8 Thomas, J. D. R., Aizawa, M., Higgins, I. J. & Albery, W. J. 1987 Immunosensors [and
Discussion]. Phil. Trans. R. Soc. Lond. B 316, 121–134. (doi:10.1098/rstb.1987.0022)
9 Aizawa, M. 1991 Principles and applications of electrochemical and optical biosensors. Anal.
Chim. Acta. 250, 249–256. (doi:10.1016/0003-2670(91)85073-2)
10 Janata, J. 1987 Chemical selectivity of field effect transistors. Sens. Actuators 12, 121–128.
(doi:10.1016/0250-6874(87)85011-4)
11 Schasfoort, R., Bergveld, P., Bomer, J., Kooyman, R. & Greve, J. 1989 Modulation of the
ISFET response by an immunological reaction. Sens. Actuators 17, 531–535. (doi:10.1016/
0250-6874(89)80041-1)
12 Schasfoort, R., Kooyman, R., Bergveld, P. & Greve, J. 1990 A new approach to immunoFET
operation. Biosens. Bioelectronics 5, 103–24. (doi:10.1016/0956-5663(90)80002-U)
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
ImmunoFETs in physiological buffer
2487
13 Bergveld, P. 1991 A critical evaluation of direct electrical protein detection methods. Biosens.
Bioelectronics 6, 55–72. (doi:10.1016/0956-5663(91)85009-L)
14 Bergveld, P., Schasfoort, R., Kooyman, R. & Greve, J. 1990 Possibilities and limitations of
direct detection of protein charges by means of an immunological field-effect transistor. Anal.
Chim. Acta 238, 323–329. (doi:10.1016/S0003-2670(00)80554-1)
15 Bergveld, P. 1996 The future of biosensors. Sens. Actuators A Phys. 56, 65–73. (doi:10.1016/
0924-4247(96)01275-7)
16 Bergveld, P. 1972 Development, operation, and application of the ion-sensitive field-effect
transistor as a tool for electrophysiology. IEEE Trans. Biomed. Eng. 19, 342–351. (doi:10.1109/
TBME.1972.324137)
17 Schöning, M. J. & Poghossian, A. 2006 Bio FEDs (field-effect devices): state-of-the-art and new
directions. Electroanalysis 18, 1893–1900. (doi:10.1002/elan.200603609)
18 Schöning, M. J. & Poghossian, A. 2002 Recent advances in biologically sensitive field-effect
transistors (BioFETs). Analyst 127, 1137–1151. (doi:10.1039/b204444g)
19 Janeway, C. A., Travers, P., Walport, M. & Capra, J. D. 1999 Immunobiology, 4th edn, pp.
79–114. London, UK: Elsevier Science.
20 Gupta, S., Elias, M., Wen, X., Shapiro, J., Brillson, L. J., Lu, W. & Lee, S. C. 2008 Detection
of clinically relevant levels of protein analyte under physiologic buffer using planar field effect
transistors. Biosens. Bioelectron. 24, 505–511. (doi:10.1016/j.bios.2008.05.011)
21 Estrela, P., Paul, D., Song, Q., Stadler, L. K. J., Wang, L., Huq, E., Davis, J. J., Ko Ferrigno,
P. & Migliorato, P. 2010 Label-free sub-picomolar protein detection with field-effect transistors.
Anal. Chem. 82, 3531–3536. (doi:10.1021/ac902554v)
22 Baumgartner, O. E., Orio, R., Wagner, P., Windbacher, T. & Selberherr, S. 2009 VISTA status
report. Vienna, Austria: Institute of Microelectronics.
23 Kang, B. et al. 2007 Prostate specific antigen detection using AlGaN/GaN high electron
mobility transistors. Appl. Phys. Lett. 91, 112106. (doi:10.1063/1.2772192)
24 Kang, B. S., Ren, F., Wang, L., Lofton, C., Tan, W. W., Pearton, S. J., Dabiran, A., Osinsky,
A. & Chow, P. P. 2005 Electrical detection of immobilized proteins with ungated AlGaN/GaN
high-electron-mobility transistors. Appl. Phys. Lett. 87, 023508. (doi:10.1063/1.1994951)
25 Kang, B. S., Wang, H. T., Ren, F. & Pearton, S. J. 2008 Electrical detection of biomaterials
using AlGaN/GaN high electron mobility transistors. J. Appl. Phys. 104, 031101. (doi:10.1063/
1.2959429)
26 Gupta, S., Wen, X., Wu, H., Nicholson III, T. R., Brillson, L. J., Wu, L. & Lee, S. C. In
preparation. Relationship between interfacial film morphology and immuno/bioFET sensitivity.
27 Nicholson III, T. R., Gupta, S., Casal, P., Wen, X., Wu, H., Brillson, L. J., Lu, W. & Lee, S. C.
In preparation. Detection of the chemokine Monokine Induced by Interferon Gamma (MIG) by
immunoFET: a tool for detection of immanent inflammatory events.
28 Wen, X., Casal, P., Lee, S. C. & Lu, W. In preparation. High sensitivity AlGaN/GaN protein
sensors with a recessed AlGaN barrier.
29 Wen, X., Gupta, S., Wang, Y., Nicholson III, T. R., Lee, S. C. & Lu, W. 2011. High sensitivity
AlGaN/GaN field effect transistor protein sensors operated in the sub-threshold regime by a
control gate electrode. Appl. Phys. Lett. 99, 043701–043704.
30 Wen, X., Schuette, M. L., Gupta, S., Nicholson III, T. R., Lee, S. C. & Lu, W. 2011
Improved sensitivity of AlGaN/GaN field effect transistor biosensors by optimized surface
functionalization. IEEE Sens. J. 11, 1726–1735. (doi:10.1109/JSEN.2010.2095458)
31 Rajpal, A., Beyaz, N., Haber, L., Cappucilli, G., Yee, H., Bhatt, R. R., Takeuchi, T., Lerner,
R. A. & Crea, R. 2005 A general method for greatly improving the affinity of antibodies by
using combinatorial libraries. Proc. Natl Acad. Sci. USA 102, 8466–8471. (doi:10.1073/pnas.
0503543102)
32 Teixeira Jr., A. L., Cardoso, F., Souza, A. L. & Teixeira, M. M. 2004 Increased serum
concentrations of monokine induced by interferon-g/CXCL9 and interferon-g-inducible protein
10/CXCL-10 in Sydenham’s chorea patients. J. Neuroimmunol. 150, 157–162. (doi:10.1016/j.
jneuroim.2004.01.013)
33 Lauw, F. N., Simpson, A. J., Prins, J. M., Deventer, S. J. V., Chaowagul, W., White, N. J. &
van der Poll, T. 2000 The CXC chemokines gamma interferon (IFN-)-inducible protein 10 and
Phil. Trans. R. Soc. A
Downloaded from http://rsta.royalsocietypublishing.org/ on June 18, 2017
2488
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
P. Casal et al.
monokine induced by IFN- are released during severe melioidosis. Infect. Immun. 68, 3888–3893.
(doi:10.1128/IAI.68.7.3888-3893.2000)
Yun, J. et al. 2002 The role of MIG/CXCL9 in cardiac allograft vasculopathy. Am. J. Pathol.
161, 1307–1313. (doi:10.1016/S0002-9440(10)64407-0)
Peprotech I. 2011 Anti-human MIG. Antigen-affinity purified polyclonal antibodies.
Systems RD. 2011 Mouse CXCL9/MIG antibody. Tools for Cell Biology Research.
Cascieri, M. A. & Springer, M. S. 2000 The chemokine/chemokine-receptor family: potential and
progress for therapeutic intervention. Curr. Opin. Chem. Biol. 4, 420–427. (doi:10.1016/S13675931(00)00113-7)
Song, J., Lu, W., Flynn, J. & Brandes, G. 2005 Pt-AlGaN/GaN Schottky diodes operated at
800◦ C for hydrogen sensing. Appl. Phys. Lett. 87, 13 3501–13 3503. (doi:10.1063/1.2058227)
Nicholson III, T. R. et al. 2009 Rational enhancement of nanobiotechnological device function
illustrated by partial optimization of a protein sensing field effect transistor. Proc. IMechE Part
N J. Nanoeng. Nanosyst. 223, 149–161.
Wen, X., Gupta, S., Nicholson III, T. R., Brillson, L., Lee, S. C. & Lu, W. 2011 AlGaN/GaN
HFET biosensors working at sub-threshold regime for sensitivity enhancement. Phys. Status
Solidi C 8, 2489–2491. (doi:10.1002pssc.201001174)
Murphy, K., Travers, P. & Walport, M. 2008 Janeway’s immunobiology, 7th edn. New York,
NY: Garland Science, Taylor & Francis Group, LLC.
Garcia-Suarez, M. d. M., Villaverde, R., Gonzalez-Rodriguez, I., Vazquez, F. & Mendez, F. J.
2009 Oriented immobilization of anti-pneumolysin tagged recombinant antibody fragments.
Curr. Microbiol. 59, 81–87. (doi:10.1007/s00284-009-9402-2)
Ha, T. H., Jung, S. O., Lee, J. M., Lee, K. Y., Lee, Y., Park, J. S. & Chung, B. H. 2007 Oriented
immobilization of antibodies with GST-fused multiple Fc-specific B-domains on a gold surface.
Anal. Chem. 79, 546–556. (doi:10.1021/ac061639+)
Torrance, L., Ziegler, A., Pittman, H., Paterson, M., Toth, R. & Eggleston, I. 2006 Oriented
immobilisation of engineered single-chain antibodies to develop biosensors for virus detection.
J. Virol. Methods 134, 164–170. (doi:10.1016/j.jviromet.2005.12.012)
Zheng, G., Patolsky, F., Cui, Y., Wang, W. U. & Lieber, C. M. 2005 Multiplexed electrical
detection of cancer markers with nanowire sensor arrays. Nat. Biotechnol. 23, 1294–1301.
(doi:10.1038/nbt1138)
Stern, E., Klemic, J. F., Routenberg, D. A., Wyrembak, P. N., Turner-Evans, D. B., Hamilton,
A. D., Lavan, D. A., Fahmy, T. M. & Reed, M. A. 2007 Label-free immunodetection
with CMOS-compatible semiconducting nanowires. Nat. Lett. 445, 519–522. (doi:10.1038/
nature05498)
Israelachvili, J. N. 1992 Intermolecular and surface forces, 2nd edn. London, UK: Academic
Press.
Gupta, S. et al. 2010 Structures and tribological properties of protein/polymer films on oxidized
AlGaN. J. Phys. D Appl. Phys. 44, 034010. (doi:10.1088/0022-3727/44/3/034010)
Bhushan, B., Kwak, K. J., Gupta, S. & Lee, S. C. 2009 Nanoscale adhesion, friction and wear
studies of biomolecules on silane polymer-coated silica and alumina-based surfaces. J. R. Soc.
Interface 6, 719–733. (doi:10.1098/rsif.2008.0398)
Phil. Trans. R. Soc. A