1. Introduction

Rapid and inexpensive detection of biomarkers has become a crucial asset in individual and global health care. Having the ability to quickly determine the health status of patients facilitates the administration of the necessary treatment by the physicians. For instance, cardiac troponin from whole blood is an early biomarker for myocardial infarctions [1]. Patients with chest pain are routinely checked for this biomarker for the correct downstream treatment. Apple et al. [2] have shown that an implementation of point-of-care testing of troponin reduces the length of stay and cost for screened versus unscreened patients. In general, biomarkers are either of physical or biochemical origin [3]. In this work we focus on the later. The detection of biomarkers is enabled by the transduction of the binding event into a readable signal. This is often performed by an electrical, electrochemical, mass-sensitive or optical transducer [4]. Preferably, the transducer works label-free, meaning that no prior sample preparations need to be performed [5]. For label-free sensing optical transducers such as surface plasmon resonance (SPR), ring resonators, slot waveguides, reflectometric interference spectroscopy (RIfS) and photonic crystal slabs (PCS) have been investigated [6–10]. Here, we use PCS.

In short, a PCS consists of a high refractive index waveguide with an incorporated nanostructured grating. Upon illumination, light that suffices the Bragg equation is coupled into the waveguide through diffraction and propagates along the waveguide as a quasi-guided mode. Due to the grating, light is coupled back out and leads to constructive and destructive interferences in reflection and transmission, respectively. This effect causes guided-mode resonance (GMR) in the spectrum [11]. The behavior is described by

In order to be used as a point-of-care device, PCS-based systems need to meet certain criteria. First, they need to be commercially feasible on a large scale, as otherwise the unit cost will be too high. This implies that the fabrication of the sensor should be cost efficient, which is achievable via mass production. Furthermore, the transportability and mobility of the setup needs to be given, as this is a core feature of a point-of-care system. In addition, a point-of-care system must be able to detect biomarkers from liquids of biological origin, such as urine, saliva or whole blood. This is of great importance as these biological samples contain the most biological information and enable an assessment of the patient’s health status, however they are also challenging to work with due to their complex nature. Preferably, the system is able to detect multiple parameters simultaneously in a multiplex fashion. Following, we briefly discuss the above imposed requirements for a PCS-based solution in the context of the current state-of-the-art and point out the existing limitations.

The nanostructures for PCS in the visible wavelength range are mainly fabricated by two processes: electron beam lithography and laser interference lithography. Current wafer-scale deep-UV lithography does not offer sufficiently small feature sizes. The resulting nanostructures may be used directly [15], or they may be replicated by nanoimprint lithography [16], or injection molding [17]. Large-scale fabrication via electron beam lithography for direct use is unfeasible due to the high fabrication cost. Hence, replication approaches need to be used for mass production. However, a replication technique is typically accompanied by the introduction of an inhomogeneity of the nanostructure across the PCS surface. Common values lie between 0.1 to 0.3 nm resonance shift per mm surface area offset [18]. Considering a 25 mm by 25 mm PCS surface, which is spotted evenly with capturing entities, the above mentioned fabrication tolerance would lead to a total change in the resonance wavelength of about 5 nm from one lateral end of the PCS to the other. For a typical resonance quality factor of Q = 100 this corresponds to the width of the resonance at 500 nm.

In order to be able to use the entire PCS surface, these fabrication tolerances need to be compensated and different solutions for this have been discussed within the scientific research community. For instance, Choi et al. [19], Zhou et al. [20], and Block et al. [21] showed possible measurement techniques for compensation. They used a spectrometer, a tunable laser, and the angular dependency of Eq. (1) combined with a monochromatic laser, respectively. However, these options do not meet the above imposed requirements. A spectrometer is usually an immovable hardware, a tunable laser is an expensive light source and the technique shown by Block et al. [21] measures in transmission, which renders the direct usage of unfiltered whole-blood unfeasible, as the solid constituents of the blood would attenuate the signal completely.

The cost, footprint and transportability is greatly improved, when combining an LED as a light source and a photodetector [22] or a camera [23]. Both works use the falling edge of the system response of their setup to transduce the resonance wavelength shift upon binding into an intensity change. The falling edge may, e.g., be caused by the LED spectrum. The resulting intensity is read out by a photodetector or camera. Due to the photodetector Lin et al. [22] are not able to measure spatially resolved, limiting a multiplex application. Jahns et al. [23] circumvented the spatial limitation by embedding the PCS between two orthogonally crossed polarizers and imaging the surface with a camera [24]. This way it is possible to track multiple parameters simultaneously. However, these setups do not compensate for different resonance positions due to fabrication tolerances, which lead to spatially dependent sensitivity [25]. In addition, both setups use a transmission mode, which excludes the use of unprocessed whole blood, as the transmitted signal would be completely attenuated.

None of the above setups meet the requirements posed at the beginning and no concepts, as to our best of knowledge, have been reported to sense biomarkers directly from whole blood with PCS. In this work we address the aforementioned and introduce a label-free tabletop measurement setup suitable for multiplex sensing and additionally demonstrate a first singleplex sensing of immunoglobulins G (IgG) from unfiltered and unprocessed whole blood based on PCS.

A sketch of the setup and its measurement concept is depicted in Fig. 1(a)). A white light source illuminates a bandpass filter. This filter is rotated to achieve time-multiplexing of the spectrum. Optical interference filters are known to blue shift their filter band when tilted with respect to a 0° angle of incidence [26]. The behavior is described by

 

Fig. 1. Overview of the measurement setup and its concept. a) Shows the underlying principle and schematic of the setup. The white light source emits a broadband light. Its spectral components are constant over time. The interference filter works as a narrow bandpass filter. Through angle-tuning the filter band is blue-shifted. This allows for wavelength filtering while maintaining the optical path nearly unchanged. A logarithmically enhanced difference image after a binding event is shown within the boundaries of the camera. b) Photograph of the setup with annotated components.

Download Full Size | PDF

The paper is structured as follows. In section 2 we introduce and describe the sensing principle and evaluation algorithm as well as the PCS fabrication, discuss design parameters, and describe the subsequent functionalization for biosensing. Results are shown in section 3 and are discussed as they are presented. Conclusions are drawn in section 4.

2. Methods and materials

2.1 Sensing principle and evaluation algorithm

A crossed polarizer setup is used here in order to only detect the resonance light and suppress the background excitation light [10,24]. The PCS is placed at an angle of 45° with respect to the polarizing axes of the two polarizers. This way, the first polarizer allows a coupling of a TM and TE mode to the waveguide. Any non-coupled light, i.e., reflected light, will be suppressed once it hits the second polarizer due to the orthogonality of the polarizing axes of the two polarizers. Only the GMR, which experiences a change in the polarization axis upon interaction with the PCS (this is a filter effect and not a polarization rotation [24]) is able to pass the second polarizer. This is the same effect as obtained when placing a third polarizer between two crossed polarizers. The camera allows for spatial resolution. Combining this with multiple spotted capturing entities allows for camera assisted multiplex detection. The PCS is illuminated by a white light source (HLV2-22SW, CCS Inc.) and the GMR light from the PCS is projected onto a camera. The light that is coupled into the waveguide as a quasi-guided mode has an evanescent field extending out of the waveguide into the cover liquid. This field decays approximately within 50 nm to 100 nm distance from the surface [17]. Any change of refractive index at the surface of the PCS will be sensed by the evanescent field and in turn changes the resonance wavelength λ_GMR according to Eq. (1). The filter (LL01-568-25) is placed onto a rotation stage (DT-80, Physik-Instrumente). By angle tuning the filter band is blue-shifted and sweeps over the resonance spectrum. The amount of light detected on the camera depends on the overlay of the filtering band with the spectrum. This leads to darker and brighter images, depending upon the angle, as is shown in Fig. 2(a)). The rotation stage is connected with an encoder, which sends a trigger signal to the camera’s GPIOs, once a certain angle has been passed. This means that after a fixed angle the camera acquires an image. During a measurement, the angular range is swept multiple times and the images are always taken at the same angles as in the sweep before. The scanning speed is approximately 0.5 nm/sec and an image is taken approximately every 0.1 degrees. A sweep has a duration of approximately 20 seconds. Local changes of the refractive index through binding lead to a local shift of the resonance wavelength. This is transduced to an angle-dependent intensity change on the camera. An example of a difference image is shown in Fig. 1(a)).

 

Fig. 2. a) The angle-tuning of the filter leads to a blue shift. At different filter angles, different wavelengths are sampled and the intensity image is taken by the camera. The images are enhanced in their contrast for a better visual. b) Rendered explosion view of the optofluidic chamber. c) Cut-out of the optofluidic chamber with its inflow and outflow cannulas, connected syringe, and waste. The cannulas pierce the gasket, enabling the liquid flow.

Download Full Size | PDF

The binding kinetics are observed by continuously sweeping the filter and image acquisition. The design and setup of the optofluidic chamber are shown in Fig. 2(b)) and Fig. 2(c)). In Fig. 2(b)) a rendering of the exploded view of the measurement chamber is illustrated. A fluid cell is formed by placing the PCS, a gasket, and a second glass plate in a stack and is sealed off by clamping them between two plates. The volume of the fluid cell is approximately 1.2 ml. The liquid is exchanged by adding at least 3 ml via a syringe and wingflow cannulas, which are injected through the gasket into the chamber cell.

The tilting effect on the filter properties is illustrated in Fig. 3. By tilting the filter the center wavelength of the transmission region is blue-shifted as predicted by Eq. (2). As seen in Fig. 3(a)), a tilt by 20° does not influence the spectral shape of the transmission region significantly. Further, a tilt by 20° leads to a blue shift of the transmission region of approximately 10 nm, as is illustrated in Fig. 3(b)). Fitting the acquired center wavelengths leads to an R²-value of 0.999, indicating an excellent fit of observed and expected behavior. Figure 3(c)) portrays the spectral response of a PCS between two crossed linear polarizers. The resonance at approximately 560 nm is the TM mode and the resonance between 620 nm and 640 nm is the TE mode. The dip of the TE mode is most likely caused by incoherent illumination [28]. The TM mode is used here as its Q-factor is higher, enabling a complete sweep of the filter transmission band over the TM mode.

 

Fig. 3. Illustration of the tilting effect on the transmission properties of an optical interference filter. a) Blue-shifting effect of an optical interference filter under angular tilting. b) Shift of the center wavelength under angular tilting. The black line is the corresponding fit. c) Guided mode resonance behavior of a photonic crystal slab under crossed polarizers. The TM mode is sampled by shifting the filtering band over the peak via angle tuning. This leads to observable changes in intensity. This is indicated by the black lines surrounding the TM mode.

Download Full Size | PDF

The evaluation process is depicted in Fig. 4. After binding, the areas of immobilized capturing entities show a change in angle-dependent intensity due to local resonance shift as visible in Fig. 4(a)). The spotting of the capturing entities is done manually (see section 2.3). This would be automated in a commercial setup. Due to the manual handling, the spots are imperfect and show irregularities. This is visible within the camera image of Fig. 4(a)) and is encountered in almost all measurements shown here. Further we want to point out that all images of binding events shown in this work are logarithmically enhanced difference images. This means that an image at the same angle at the start and at the end of the measurement are subtracted from one another and a logarithm is applied to this difference image. The logarithm enhances small intensity changes and facilitates visual understanding. Due to the irregularities we define regions of interest (ROIs) for the immobilized entities (blue) and their corresponding references (black) manually as seen in Fig. 4(b)). The pairing of spots and reference are customizable. Each spot and reference area is connected via a line with graded coloring indicating their pairing. In general, the reference areas are not functionalized with capturing entities, but simply saturated with bovine serum albumin (BSA) (see section 2.3). By referencing to these saturated areas, the drifts induced by surface fouling, non-specific binding, or temperature changes are compensated. The difference signal is then attributed to the specific binding of the analyte to the capturing entity.

 

Fig. 4. Concept of the evaluation algorithm. a) Logarithmically enhanced difference image after a multiplexed binding event. b) Regions of interest for functionalized areas in blue and their corresponding reference areas in black for the regions of interest in a). The pair of areas (spot and reference) are connected via a graded color line, in order to illustrate their pairing dependency. c) Normalized intensity sweep for a functionalized area in blue and its corresponding reference in black. The encoder difference at 80% of the maximum value is tracked and shown in red in the inlet.

Download Full Size | PDF

After placing the ROIs for the spot and reference region, each sweep is analyzed for its intensity within these ROIs. For each angle position, the algorithm calculates the average intensity within the given ROIs. By doing this for every image within a sweep, it is possible to reconstruct the spectrum of the PCS. In our case the TM mode of the PCS. This is shown in Fig. 4(c)). It illustrates the cropped tail of the TM mode. Following, the algorithm detects the maximum of a sweep by parabolic fitting and normalizes all images of a sweep to this maximum intensity as seen in Fig. 4(c)). A linear fit of the falling edge at the shorter wavelength is applied. The 80% value of the maximum is tracked for the binding and referencing areas. Due to binding the two intensity profiles are shifted differently and by calculating the difference of the 80% value the signal change is recorded. The red bar indicates that level. It is measured in encoder shift units. This allows for on-chip compensation, such as drift, unspecific binding or fouling [29]. We want to highlight that each measurement point consists of a sample and a referencing area. This is also true for reference areas. For this case two background ROIs are compared to one another, leading to the absence or reduction of a signal, as we compare two areas, which are saturated with BSA. This evaluation method was entered as a patent [30]. Due to the manual curation it takes about 10 minutes after the end of the experiment to obtain the graphs. As mentioned above, the measurement is performed in reflection to mitigate optical disturbances and allows to measure opaque liquids.

2.2 Fabrication of photonic crystal slabs

PCS are fabricated using nanoimprint lithography as depicted in Fig. 5 [23]. A nanostructure with a period length of 370 nm, and a depth of 45 nm is used as a glass master (custom product, AMO GmbH). A negative soft mold of the glass master is formed by pouring poly(dimethylsiloxane) (PDMS) (Sylgard 184, Dow) onto the master structure. The PDMS is mixed in a ratio of 8:1 with its curing agent.

 

Fig. 5. a) Schematic of the nanoimprint lithography process for the fabrication of photonic crystal slabs. b) An SEM image of the photonic crystal slab surface. Prior to measurement 10 nm gold were sputtered onto the titanium dioxide (TiO₂) layer for a better contrast. c) Corresponding resonance spectrum of the photonic crystal slab in b) with water as a cover medium. Spectrum is taken between two crossed linear polarizers.

Download Full Size | PDF

The PDMS and glass master are placed in an oven for 30 minutes at 130° C. After cooling, the PDMS is peeled off the glass master and the PDMS stamps are cut into squares of approximately 25 mm x 25 mm in size. We imprint the stamp into a spincoated UV nanoimprint resist (Amonil, AMO GmbH). The stack is UV cured for 1 minute and the PDMS is peeled off, exposing a nanostructured resist. Last, we sputter a high refractive index layer of 55 nm titanium dioxide (TiO₂) onto the nanostructure as the waveguide layer. This leads to a TM mode at about 560 nm. In Fig. 5(b)) an SEM image of a PCS surface is shown. For better contrast, 10 nm gold were sputtered onto the titanium dioxide layer prior to SEM imaging. The reproduced nanostructure is clearly visible. The period length is measured to be 365 nm, which corresponds to a shrinkage of approximately 1.4%. PDMS is known to exhibit a shrinking behavior, when cured [31]. The corresponding resonance behavior under crossed polarizers is shown in Fig. 5(c)).

Via sputtering of the waveguide layer it is possible to tailor the resonance position of the TM or TE modes, as the thickness of the guiding layer directly influences the effective refractive index and by it the resonance wavelength [14]. The TM mode is chosen here, as it is a bit more sensitive towards refractive index changes at the surface [14]. By using 370 nm as the period length we are able to tailor the TM mode to be in the green color regime, which is usually the most sensitive color channel of a camera. The depth of the nanostructure influences the Q-factor of the resonance [32]. The shallower the grooves, the higher the Q-factor. However, shallow grooves lead to less scattering and as a consequence to less light detected with the camera. A grating depth of 45 nm is a good trade-off between resonance finesse and light detection, which is used for the PCS fabrication in this work.

2.3 Surface functionalization and immobilization of capturing entities

In order to immobilize capturing entities on the surface of the PCS a functionalization using click-chemistry is used [33]. For functionalization two routes are being used. The first one is based upon vapor-phase deposition and the second one exposes the PCS to liquids to achieve surface modification. The second route has been described by Jahns et al. [23]. Here, we limit ourselves to the detailed description of the vapor-phase process and refer to the processes as route 1 and 2 for the vapor-phase and liquid-phase based approach, respectively.

The PCS is cleaned in an ultrasonic bath for 10 minutes with acetone and isopropanol, consecutively. Next, the surface is activated by placing the PCS in a reactive ion etching chamber (Sentec, SI100) for 3.5 minutes with 100 W power and an oxygen flow of 8 sccm. Following the activation, the PCS is placed on top of a small beaker, facing with the activated surface down. The beaker contains 400 µl silanization solution, which consists of 260 µl of (3-Aminopropyl)triethoxysilane (APTES) (440140, Sigma-Aldrich) dissolved in 24.5 ml of dry methanol (322415, Sigma- Aldrich). In order to facilitate evaporation, we place the beaker and PCS in a desiccator and connect it to the inhouse vacuum pump. The PCS is left exposed to the silanization vapor for 1 hour. Afterwards, we place the PCS on a hotplate of 110° C to desorb excess silanization molecules for 20 minutes. Next, the crosslinking step follows. Here, we dissolve 200 mg of 1,4-phenyldiisothiocyanat (PDITC) (258555, Sigma-Aldrich) in 1 ml of pyridine (270970, Sigma-Aldrich) and 9 ml N,N-dimethylformamid (DMF) (227056, Sigma-Aldrich). Once more, we place the silanized PCS face down on top of a beaker, which contains the crosslinking solution. Analogue to step one, the beaker and PCS are placed into a desiccator and the desiccator is connected to the inhouse vacuum pump. Due to the slower rate of evaporation this step is left running for 3 hours. Afterwards, we place the crosslinked PCS in a fume hood for 20 minutes to let excess crosslinker desorb.

The capturing proteins are immobilized by placing drops of 1 µl on the PCS manually, which are left for overnight incubation in a wet cell. The next day the remaining surface is passivated by a concentration of 1 mg/ml of bovine serum albumin (BSA) (05470, Sigma Aldrich) in Dulbecco’s phosphate buffer (DPBS) (D8537, Sigma Aldrich). After 1 h the BSA is rinsed off with DPBS and the surface is cleaned with deionized water and dried with nitrogen. The passivated area acts as the reference area during evaluation.

Different capturing entities were used. The GST antigen (custom made, Novatec Immundiagnostica GmbH) was spotted at a concentration of 100 µg/ml and is used for the detection of GST-antibodies (G7781, Sigma Aldrich). This corresponds to maximum immobilized mass of 100 ng. For the detection of streptavidin (S0677, Sigma Aldrich) BSA conjugated biotin (A8549, Sigma Aldrich) is used and spotted at concentrations of 100 µg/ml and 10 µg/ml, which corresponds to a theoretical limit of 100 ng and 10 ng. For the detection of thrombin (HCT-0020-MG, CellSystems GmbH) a custom made aptamer is spotted at a concentration of 10 µM, which corresponds to a theoretical immobilized mass of 92 ng. The nucleic acid strain of the aptamer is 5’-C6-Aminolink-XAG TCC GTG GTA GGG CAG GTT GGG GTG ACT-3’, with X = Spacer-HEG/PEO (custom made, metabion international AG). The detection of antibodies from whole blood is enabled by immobilizing protein A/G (21186, ThermoFisher) at a concentration of 100 µg/ml on the surface. Similarly, this corresponds to a maximum immobilization mass of 100 ng. Protein A/G is a recombinant protein, which binds antibodies of type IgG with high and antibodies of type IgA and IgM with low affinity [34]. All samples are diluted in DPBS.

3. Results

3.1 Performance evaluation and singleplex detection of anti-GST

First, the ability of the setup to detect changes of bulk refractive index and singleplex detection is investigated. The detection of bulk refractive index changes acts as a proxy of its general performance. The results are shown in Fig. 6. In order to have areas, which are not affected by the change of bulk refractive index, we add PDMS drops on the PCS surface. The PDMS seals off the surface of the PCS at those points. Now, when the content of the optofluidic chamber is changed, the change in refractive index will not be detected for the areas under the PDMS. These areas are therefore used as the reference areas for the evaluation algorithm. Comparing bare areas with the sealed off areas allows to investigate the sensitivity of the PCS. This is shown in the image inlet in Fig. 6(a)). In total, 4 areas are evaluated. The pairing of the sensing and referencing area is shown by a graded color line.

 

Fig. 6. a) Performance evaluation via a refractive index sweep between water and 5% glycerin. The image inlet shows four PDMS spots, which were used as a sealed off reference. b) Recorded change in encoder position after exposure to GST antibodies at a concentration of 1:250. The inlet is a logarithmically enhanced difference image between the start and end points and shows the 4 positions of binding. The two black spots are the reference area pair. The blue shaded area indicates the one sigma interval. Initially DPBS was in the fluid cell. At approximately 8 minutes the 1:250 diluted GST antibodies were injected into the fluid cell. At approximately 32 minutes DPBS was added again.

Download Full Size | PDF

The optofluidic chamber is initially filled with water. After 5 minutes, a mix of 5% glycerin in water is added and after another 5 minutes the fluid cell is flushed with water again. The recorded (Krüss, DR-201-95) refractive indices are 1.3331 and 1.3392 for water and 5% glycerin, respectively. The average step height of all four PDMS drops is 170 encoder shift units. The corresponding sensitivity is calculated by dividing the step height by the refractive index change. The resulting sensitivity is approximately 28000 encoder shift units per refractive index unit. The limit of detection is calculated by dividing the triple noise level by the sensitivity. The triple noise level is 9 encoder shift units, which leads to a limit of detection of about 3.4 E-4 refractive index units. This is comparable to the work published by Lin et al. [22]. However, the signal exhibits a slight drift when exposed to the glycerin step. We assume that liquid enters the resist through cracks in the PCS surface, as we used this effect previously to fabricate flexible PCS [35]. This might explain the observed offset after the second exposure to water, as well.

The biosensing functionality of the setup is demonstrated by detecting GST-antibodies diluted by 1:250 in DPBS on a PCS functionalized by route 2 with 4 spots of the GST antigen with a concentration of 100 µg/ml (100 ng). GST-antibodies are used, since they are thoroughly researched and are employed as tags in the purification of recombinantly expressed proteins [36]. This makes them an ideal choice to investigate proof of principle interactions and in addition they were readily available. We want to note here that the manufacturer does not supply an exact GST-antibody concentration, as the product is a polyclonal antibody. The manufacturer only states the overall IgG concentration range, which lies between 8 mg/ml to 15 mg/ml. Since we cannot calculate an exact concentration, we only state the dilution factor and not a resulting concentration. The results are shown in Fig. 6(b)). After a baseline phase of 8 minutes with DPBS the GST-antibodies diluted by a factor of 250 are injected. At 32 minutes the fluid cell is washed with DPBS again. The inlet shows four spots, where the binding was detected in blue and their corresponding reference area in black. They are connected via a graded color line. The behavior of the area without any capturing entities is shown by two black ROIs connected by a black line. The detected signal change is approximately 32 encoder shift units. The average noise is approximately 5 encoder shift units. The signal of the anti-GST is not reducing during washing. This means that the change of surface mass for the reference and capturing entity area is the same and no difference is obtained. This indicates that little unspecific binding to the saturated areas is taking place and that no excess binding of GST antibodies to the antigen is taking place, as they would be washed off during the washing step. In contrast to the black reference line a clear trend is visible; thus proving the ability of biosensing. The measurement was stopped after 32 minutes as a clear binding was visible. However, Fig. 6(b)) shows that no saturation of the anti-GST antibodies was detected. This indicates that the capturing entities are not yet completely saturated. Yet, for a functionality proof the detection of binding is sufficient.

3.2 Multiplex detection of thrombin, anti-GST and streptavidin

Next, a sample functionalized by route 1 with multiple capturing entities is investigated. The capturing entities are two drops of thrombin binding aptamers at a concentration of 10 µM (97 ng), two drops of GST-antigen at a concentration of 100 µg/ml (100 ng) and two drops of biotin-BSA with a concentration of 100 µg/ml (100 ng) and 10 µg/ml (10 ng), each. Similarly to the singleplex measurement, a baseline with DPBS is acquired for 3 minutes. Then we injected 6.3 µg/ml (87 nM) thrombin for 25 minutes. Following a 3 minute DPBS flush, GST-antibodies at a dilution of 1:250 were injected and left inside the fluid chamber for 30 minutes. After that we flushed again for 3 minutes with DPBS and added streptavidin at a concentration of 33 µg/ml (550 nM) for 5 minutes and did a final DPBS flush for 3 minutes. Figure 7 shows the results of the multiplex measurement for each immobilized capturing entity separately. The top, middle, and bottom row show the detected signals for immobilized aptamers, GST antigens, and BSA-biotin, respectively. The colors of the spots on the images on the right-hand side correspond to the colors in the plots. They are connected by a graded color line to their corresponding reference area. These reference areas are saturated with BSA and thus only unspecific binding behavior is detected. The two black spots are the reference data pair within the plot. The reference signal is the results of the comparison of two BSA saturated areas. This is the reason, why no permanent signal change is detected.

 

Fig. 7. Recorded encoder value shifts for a multiplex measurement with (top) immobilized aptamers for thrombin detection, (middle) GST-antigen for GST-antibody detection and (bottom) BSA-Biotin for streptavidin detection. The color coded spots in the inlet images correspond to the plotlines in the graphs. The corresponding reference areas are black. They are connected via a graded color line to their respective reference area. All three images are logarithmically enhanced difference images.

Download Full Size | PDF

After the injection of thrombin a binding to the aptamers is detected, as is visible from the two distinct spots in the image. The maximum recorded encoder shift lies between 40 and 20, for spot 1 and 2, respectively and the one sigma noise interval is approximately 3 encoder shift units for both measurements. A decrease in signal is detected at the washing step and it corresponds to different mass changes at the surface for the reference and capturing areas. We assume that parts of the thrombin are loosely bound to the aptamer and are detached during the washing step. This leads to a decrease in signal as the change in surface mass at the aptamer is bigger than the change of surface mass in the referenced area. The detected signal discrepancy of about 20 encoder shift units between aptamer 1 and 2 might be attributed to the manual spotting performed in these experiments. Even though the same concentration of capturing entity was spotted, it does not necessarily mean that the same mass was immobilized on the surface [37]. The discrepancy of the detected signal strength might be attributed to differences in the amount of immobilized aptamers on the surface. After the injection of DPBS, the blue signal of aptamer 1 decreases by about 20 encoder shift units and remains on that level. The orange signal for aptamer 2 exhibits a drift, which indicates desorption of thrombin from the aptamers. This effect is most likely caused by an interplay of several factors. A possible explanation is the change in conformation of the aptamer 2 after injection of the subsequent liquid.

Thrombin is a factor which regulates the clotting of blood. In unadulterated whole blood at 37°C the formation of fibrin clots occur at thrombin concentration of 10-30 nM [38]. The here detected signal of 87 nM lies above this range, yet assuming a linear relationship between analyte concentration and signal strength, a level of 10 nM to 30 nM thrombin might already be detectable, as this would lead to a signal of about 4 to 10 encoder units shift, which is slightly above the noise floor. In addition, we believe that by continued development the signal strength is improvable. For instance, different concentration levels of capturing entities affect the achievable signal strength. This effect is visible in the streptavidin binding of Fig. 7 and is discussed further below.

However, the levels of thrombin in whole blood give essential information for trauma patients. Cardenas [39] has reported that shock patients, who exceed thrombin levels above 250 nM in their blood are more likely to receive a blood transfusion and exhibit a 3-fold increase in their risk of a 30-day mortality. Our reported detection signal would be able to discriminate these levels. Further, Coleman et al. [40] have stated that the evaluation of thrombin in whole blood is of great interest for medical research, as the thrombin levels fluctuate after a severe trauma. Yet, few point-of-care devices regarding the ability to detect thrombin from whole blood have been shown. The here shown ability to detect thrombin are groundwork for a possible combination with whole blood.

The injection of GST-antibodies leads to an encoder shift of approximately 80 and 57 encoder shift units for spot 3 and 4, respectively, and a detected one sigma noise level of approximately 4 and 3 encoder shift units, respectively. The discrepancy of the signal might be attributed to the same effect as for the thrombin binding. Again, the measurement is stopped as a clear signal is visible.

Last, the binding of streptavidin was investigated. The two spots exhibit a signal change of approximately 420 and 180 encoder shift units for spots 5 and 6 after being exposed to a concentration of 33 µg/ml, respectively. Spot 6 was immobilized with 100 µg/ml and spot 5 with 10 µg/ml. The detected one sigma noise levels are 11 and 9 for spot 6 and 5, respectively. In this case, the difference in signal is attributable to the difference of available capturing entities on the surface. Due to the ability to resolve spatially, it would be possible to use this setup for one-shot titration experiments [41]. This could greatly reduce time and costs for the fabrication of test kits. All three binding events show that no cross-talk or unspecific binding is occurring. Furthermore, the three binding events indicate that the detection of thrombin and streptavidin is happening in excess of available proteins, as seen by their signal half-time of about 30 seconds, whereas the binding of the GST-antibodies is slower. The signal half-time is approximately 10 minutes. It is to be mentioned that the binding affinity is also influenced by the amount of available ligand. In order to improve the binding kinetics further titration experiments need to be performed [42]. Also repeatability tests for the multiplex binding event shown here need to be carried out in order to quantify the specifications. However, this is beyond the scope of this work here and should be addressed in future work.

3.3 Detection of immunoglobulins G (IgG) from whole blood

Last, we show the detection of immunoglobulin G (IgG) antibodies from undiluted and unfiltered whole blood. It contains all the constituents of blood and is therefore an opaque liquid. The results of this measurement are shown in Fig. 8. The blood was obtained in the morning of the measurement from one patient in the university hospital of Schleswig-Holstein (UKSH) with citric acid as the anticoagulant. It was stored in a refrigerator and was taken out 10 minutes prior to the start of the measurement. No additional processing of the blood was done pre measurement. In total, three blood samples were taken and measured each on a different PCS. The blood samples were taken according the ethical framework given by the ethics commission of the UKSH with the number D444/22. Consent was given by the patient prior to the blood collection. Each PCS sample was functionalized by route 2 and spotted with five 1 µl drops of 100 µg/ml protein A/G (100 ng). In Fig. 8(a)) the detected encoder shifts for each sample are illustrated. The inlet shows the logarithmically enhanced intensity difference after the exposure to blood. Each graph shows the average value of the five spots with their one sigma noise level in a shaded color. The active and background areas are again indicated by blue and black spots connected by a graded color line and the reference area pair is shown in black.

 

Fig. 8. Whole blood sensing. a) Detected encoder shifts for three different samples. All three PCS samples were all immobilized with five 1 µl drops of 100 µg/ml protein A/G (100 ng). After 5 minutes of DPBS exposure whole blood taken from a patient was injected. After another 10 minutes the fluid chamber was washed with DPBS. The reference spots are black and the capturing spot are blue. The reference is measured by comparing two background areas, shown here in two black spots connected by a black line. A pair is connected via a line with graded coloring. The images are logarithmically enhanced difference images. Whole blood means that no filtering or processing step prior to measuring was executed. Blood was stored in a refrigerator and was taken out 10 minutes prior to the start of the measurement b) Image of the inflow and outflow cannulas filled with blood. c) Fluid chamber after the final washing step. The lower left corner shows the opaqueness of the blood.

Download Full Size | PDF

All three samples exhibit a distinct binding signal with values between 450 and 600 encoder shift units. The one sigma interval for the first two measurements lies at approximately 50 encoder shift counts. The third measurement has a one sigma value of approximately 100 encoder shift units. This behavior might be attributed to the fact that this measurement was taken in the afternoon, whereas the first two samples were measured in the morning. Blood is a highly complex medium and it is known to be affected by storage time and conditions [43]. For all three measurements, the detected signal half-time is less than 20 seconds, which is similar to the thrombin and streptavidin measurement from Fig. 7 in section 3.2 and indicates an excess of binding material. However, compared to the obtained signals in Fig. 6 and Fig. 7 for GST-antibodies and thrombin, the detected changes are a lot stronger for the whole blood measurement. This is attributable to the abundance of antibodies within blood and the high affinity of protein A/G to bind these. Only the binding of streptavidin to biotin shows a similar signal change. Biotin and streptavidin are known to have a strong binding affinity [44]. However, due to the abundance of biological material a background signal change is detected. In measurement one and two a clear signal change of the reference is detected. This indicates that unspecific binding and fouling is taking place at the surface. This effect might limit the detection of lower concentration biomarkers, such as thrombin, directly from whole blood. Kingshott and Griesser [45] have shown that unspecific binding is minimized when poly(ethylene glycol) (PEG) is introduced at the surface. We have recently shown, that we are able to functionalize PEG containing molecules to a glass surface and are able immobilize aptamers to it [46]. We aim to use this functionalization route for specific biomarker detection from whole blood in order to minimize fouling effects. Furthermore, Fig. 8(b)) illustrates the inflow and outflow of the blood and Fig. 8(c)) depicts the remains of the whole blood in the fluid chamber after the washing step. The lower left corner shows the opaqueness of whole blood. Only through measurement in reflection is it possible to detect binding from whole blood. These results show that the solid constituents of the whole blood do not hinder label-free biosensing from whole blood, however care needs to be taken when aiming to detect biomarkers with a lower concentration. This might be attributable to the surface sensitivity of the PCS, which probes up to approximately 100 nm above the PCS surface [17].

Specific subtypes of IgG antibodies are of interest for medical applications. Their concentrations are indicative of an acute disease the patient is experiencing. For instance, Brokstad et al. [47] have shown that after a vaccination with influenza A and B strains, patients exhibit an increased level of IgG after 8 days after the vaccination. Detecting specific IgGs allows to evaluate whether a patient has been exposed to a pathogen. A prominent example is the antibody detection of Sars-CoV-2 [48]. Even though we detect all types of IgG subtypes when using protein A/G as a capturing entity, we do believe that the here shown detection limits enable the detection of specific subtypes of IgG. In a recent publication, we have shown that we were able to detect IgG antibodies binding to immobilized protein A/G from diluted feline serum at a dilution factor of 500 [35]. Since the functionalization is done the same way in this work, we expect that we would be able to detect IgG antibodies, if we were to dilute the whole blood by the same factor. This would corresponds to a detected mass concentration at the surface of 10-100 µg/ml. IgGs are also used for assessment of autoimmune diseases as well as cancer.

While the here shown detection levels in the range of µg/ml are quite high for various biomarkers, possible application fields are thinkable. Khan et al. [49] have recorded concentration levels of IgG in human blood in the range of 0.86 to 4.76 mg/ml. Setting the obtained levels of proteins in relation to one another, one can argue that we are able to detect mass concentrations from the low microgram per milliliter range to the milligram per milliliter range. This is a relevant concentration range for trauma patients, when they are checked for fibrin degradation products (FDP) and d-dimer (DD) levels. Hagiwara et al. [50] have shown that both biomarkers are indicative of the trauma level a patient has experienced and they range from 5.7–160 µg/ml for FDP and 3.2–240 µg/ml for DD. Further, as stated in section 3.2 thrombin is an essential biomarker for polytrauma patients. The here reported detection levels could be able to assess the injury level of trauma patient quickly, as the whole blood is directly applicable.

4. Conclusions

In this work we have shown for the first time the ability to detect biomarkers from unprocessed and unfiltered whole blood with photonic crystal slabs. For this, we have devised a new reflection based measurement setup, which uses an angle-tuned filter for label-free sensing. The reflective setup allows for direct measurement of unprocessed whole blood, as the solid constituents do not perturb the detection. In addition, we were able to show a multiplex platform with immobilized single DNA strands (aptamers), antigens, and proteins. This elucidates the platform’s strength for a wide range of multiplex applications. The setup itself expands the detection methods known from the scientific literature for PCS sensing. Furthermore, the angle tuning enables the compensation of fabrication inhomogeneities. This reduces tolerance requirements for the fabrication process and paves the way to mass fabrication by nanoimprint lithography or injection molding. Our evaluation algorithm allows for continuous difference measurements between binding sites and reference sites revealing the binding kinetics. The setup’s footprint and optomechanical requirements are moderate, showcasing the ability to be used as a transportable point-of-care setup.

The current status of the setup demonstrates its potential for point-of-care testing well, but also points out areas, which need to be addressed by further research and development. The performance of the setup is greatly influenced by encoder resolution, rotation speed and amount of images acquired per sweep. We expect an improvement of the overall performance by refining these parameters and their interplay. In addition, Li et al. [51] have shown that the choice of the fitting algorithm strongly affects the achievable LOD. We expect an improvement within this field with according research, as well. The observed limit of detection of 3.4 E-4 RIU is of comparable performance to other measurement setups [22], but not as good as the one shown by Li et al. [51]. However, we show that we are able to sense from unprocessed whole blood, which is a strong advantage for a point-of-care setup.

After further developments and reaching commercialization, the cost per test will include the cost of the sensor system, the cost of the functionalized PCS sensor, and the medical personnel cost. The sensor system itself can easily be realized as a compact system that can be well cleaned. Due to the remote optical sensing no contamination occurs during measurement. The fabrication cost of the PCS chip is very low when mass production by injection molding is reached (similar to DVD fabrication). The sensor cost will be dominated by the surface functionalization with capture molecules. The functionalized PCS chip is a one-way product. The PCS itself can be recycled after cleaning. We typically reuse the PCS many times after oxygen plasma cleaning. The final cost per test will likely be dominated by the medical personnel cost. The approach will be particularly beneficial for a high degree of multiplexing. With our approach spotting an array of 10 × 10 is easily achievable. Based on the observed reaction times we expect that the results will be available approximately 5 minutes after inserting the whole blood. Robotic handling of samples can also be well integrated with our approach.

Finally, the illustrated detection range of biomarkers from 6.3 µg/ml to 2 mg/ml is already within the mass concentration range of clinically relevant biomarkers such as FDP, DD and thrombin and could allow for direct evaluation of injuries in shock patients from unprocessed whole blood. This needs to be investigated in further research.