A nanofluidic device for rapid and multiplexed SARS-CoV-2 serological antibody detection


 The outbreak of COVID-19 has led to a substantial death toll and has hindered the functioning of modern society, sending the world into a medical and economic crisis1,2. This underlined the importance of point-of-care diagnostics, as well as accurate, cost-effective serological antibody tests as well as point-of-care diagnostics to monitor the viral spread and contain pandemics and endemics. Here, we present a three-dimensional (3D) nanofluidic device for rapid and multiplexed detection of viral antibodies. The device is designed to size-dependently immobilize particles from a multi-particle mixture at predefined positions in nanochannels through capillary forces only, resulting in distinct trapping lines. We show that individual lines can be used as an on-chip fluorescence-linked immunosorbent assay for multiplexed detection of serological immunoglobulin antibodies against viral proteins with high sensitivity. Further device versatility is exhibited by on-bead color multiplexing for simultaneous detection of IgG and IgM antibodies in convalescent human serum and by concurrent detection of anti-spike (SARS-CoV-2) and anti-hemagglutinin (Influenza A) antibodies. The device’s applications can be further extended to detect a plethora of diseases simultaneously in a reliable and straightforward manner.

Our nano uidic device is designed to be low-cost and easy to operate. It is entirely made of PMMA and contains only passive capillary micro uidic elements to control the uid ow without the need for external loading mechanisms, such as syringes or pressure pumps. The critical feature of the device is a wedge that enables geometrical trapping of nanoparticles ( Fig. 1a and j) at a predetermined location, allowing for particle size sorting and size determination (Table S1).
The chip's in ow region consists of three parallel channels with separate inlets (three channels in the present case) to enable redundancy in testing or multiplexing (Fig. 1f). The applied liquid is passively aspirated into the 300-µm-wide in ow, enhanced pillars with a 20 µm diameter and 40 µm pitch (Fig. 1c).
The channels containing various passive ow elements were optimized for a controlled and reproducible ow. More speci cally, the size and pitch of the incorporated pillars were altered throughout the device ( Fig. 1c-e) to modulate uid ow rate and to provide necessary support during the bonding of these relatively broad, but extremely shallow channels (aspect ratio 375:1) (Fig. S.2). Furthermore, to minimize premature drying of the liquid inside the nano uidic device and to ensure a constant uid lling velocity, an in ow resistor was incorporated (Fig. 1b). This was achieved by narrowing the channel width after the in ow region to 100 µm and keeping it constant for a length of 500 µm. Afterward, the width is gradually increased to 300 µm ( Fig. 1b -I.R.). The IR is followed by the trapping region (TR), i.e. a wedge-structure with a linearly decreasing the channel height from 3.4 to 0.8 µm over a distance of 1.2 mm.
After the TR, the channel height is kept constant at 0.8 µm for 100 µm before a secondary wedge increases the total height to 1 µm over 1 mm. This connecting wedge (CW) couples the active device region to a capillary pump containing 20 µm pillars with a 30 µm pitch (Fig. 1e). The TR and CW necessitate nanoscale topography control over several millimeters and simultaneous integration of various micro and nano uidic components. To overcome these challenging fabrication requirements, high precision gray scale e-beam lithography (g-EBL) is employed to pattern the nano uidic device following a process that we developed recently through extensive process optimization 12 . This highresolution structure is subsequently replica-molded to obtain a negative polymeric stamp, enabling costeffective upscaling of fabrication through nanoimprint lithography (Fig. S1).
The topography variation in the TR is linear and patterned with high accuracy (Fig. 1g-h). The interference colors ( Fig. 1h -inset) visible after bonding evidence the preservation of the 3D pro le inside the nano uidic channel. Before proceeding with the particle trapping experiments, the devices were functionalized with poly(vinyl alcohol) (PVA). The application procedure of PVA was shown to in uence the ow behavior of the particles in the nano uidic channels and was extensively optimized to maintain a high owrate during the entire particle trapping experiment (Sup. Video 1). The functionalization blocks non-speci c binding sites on the PMMA 13 (Fig. S3) and allows for more controlled surface wetting 14 . The latter is evidenced by the absence of corner ow in the capillary pump region and improvement of the lling front (Fig 1i; Fig. S4; Sup. Video 2). The device concept with various micro and nano uidic components in combination with speci c surface chemistry allows for facile device operation that can be used in various applications. This includes particle size determination, sorting, and immobilization, as well as paving the way for applications such as in point-of-care diagnostics.
The ease-of-use and the performance of the device was highlighted by passive size-dependent particle immobilization of ve different uorescent polystyrene calibration grade particles in the TR (Fig. 1 j-k). The relationship between the particle trapping position and its nominal diameter was shown to be highly linear (R 2 = 0.998; Fig. 1l) and could be used to effortlessly determine particle sizes with nanometer accuracy (Table S1). Extrapolation from the linear t (Fig. 1j), amounts to a minimal TR channel height of 720 nm, indicating a reduction of roughly 45 nm when compared to the unbonded height of 765 nm. This can be attributed to the surface selective bonding process [15][16][17] and also highlights the necessity of supporting pillars inside the nano uidic channel (Fig. 1h, inset; Fig. S2).
By using the size-dependent immobilization properties of our nano uidic device, we developed an onbead immunoassay for the detection of speci c SARS-CoV-2 antibodies in human serum by concentrating the relevant nanoparticles at predetermined locations. Our immune system ghts infection and eliminates foreign bodies using innate and adaptive mechanisms, where the latter requires growth and rearrangement of gene elements to produce antibodies that speci cally bind to invading antigens.
The most commonly found antibody type in human serum is immunoglobulin G (IgG) 18 , making it a valuable target for immune status characterization. One of the major antigens of SARS-CoV-2 associated IgG antibodies is the surface trimeric spike (S) protein, which plays a vital role in facilitating the entry of the SARS-CoV-2 into human cells 19 . The receptor-binding domain (RBD) of the S-protein (S-RBD) is an immunodominant target for SARS-CoV-2 antibodies 20 . To enable their on-chip detection, we used the high-a nity interaction between streptavidin and biotin to couple biotinylated S-RBD to the surface of 2.8 µm streptavidin-coated beads ( Fig. 2b; top).
Once functionalized, the beads were mixed with human serum. If the patient was previously exposed to SARS-CoV-2, the serum will contain antibodies with a speci c antigen-binding site for S-RBD and bind to the beads. The latter can be uorescently visualized by conventional immunostaining procedures. In our case, a donkey-originating antibody conjugated with a red uorescent dye (Cy5) was used that speci cally binds to human IgG ( Fig. 2a; left). To benchmark the serological immunoassay, the lowest detectable concentration of a puri ed humanized anti-S-RBD IgG antibody in human serum was determined. To evaluate the nano uidic device function, non-functionalized 0.9 µm calibration-grade particles were added to the suspension, acting as a ow control before applying the particle mixture onto the device's in ow region. Analysis of the corresponding uorescent signal of the trapped beads revealed a limit-of-detection (LOD) around 1 nM (Fig. 2c). This sensitivity is comparable to the state-of-the-art micro uidic devices 10,21,22 , whereas the present device is more straightforward to use and has a low infrastructure requirement. Furthermore, in recent studies, the physiologically relevant concentration of this speci c type of anti-spike IgG antibody in recovered COVID-19 infected patients has been shown to be in the range of 9.6 -28 880 nM 23 . This shows that the developed on-bead and on-chip nano uidic device operates well within the physiologically relevant concentrations.
Additionally, the assay was validated using a patient serum set containing both PCR positive (n = 19) and PCR negative (n = 10) patients (Fig. 2d). From the 29 different samples, 28 could be identi ed correctly according to infection status, with one false negative result. The signal distribution of positive cases is to be expected, as immune responses are highly individual and dependent on the infection history of the patient. This corresponds to a test sensitivity of 94.7% (99-65) with a speci city of 100% (100-72) and an area under the curve of 0.95 (σ x̅ = 0.05) according to receiver operator characteristic (ROC) analysis. We note that the false negative was retested and showed a signal above the threshold value. Although we could not reconstruct the root cause of the initial false negative during the blind experimental run, the overall results show the performance of the devices in this early development phase and highlights the potential for further optimization. These ndings evidence the applicability of the assay in a real-world setting for the IgG serological characterization of suspected COVID-19 patients 24 .
Even though IgG is the dominant immunoglobulin subtype in human serum, it is mainly associated with late-stage and memory-related immune responses. In the case of SARS-CoV-2 infection, speci c IgG antibodies become prominent within 7 days post-infection. This hampers the applicability of IgG-speci c antigen assays for early-stage disease detection. However, other antibody sub-types, such as immunoglobulin M (IgM), can be already found in human serum 4 days post infection 25 . Hence, the onbead and on-chip assay was further developed to include color-multiplexing for simultaneous detection of both antibody subtypes ( Fig. 2e; Fig. S5). More speci cally, the immunostaining solution contained anti-human IgM antibodies conjugated with a green uorescent dye (Alexa 488) as well as anti-human IgG antibodies conjugated with a red uorescent dye (Cy5). The anti-IgG uorescent signal of the COVID-19 positive patient was 25.5 times greater than that of the negative control. Similarly, the anti-IgM signal was 4.9 times higher (Fig. 2f). This difference between IgG and IgM is to be expected, given that the investigated COVID-19 positive serum was taken 33 days after symptom onset. As is well known, IgM antibody levels start to decline roughly 21 days after infection, explaining the ndings 26 . This experiment shows the potential of color multiplexing to monitor the IgG and IgM response of suspected SARS-CoV-2 infected people.
Furthermore, it is of interest to perform simultaneous differential testing for the presence of antibodies against diseases with similar symptoms. In the case of COVID-19, one of the most symptomatically similar and prevalent viral infections is In uenza A. The latter causes yearly epidemics and it is one of the major targets of annual vaccination campaigns 27 . To demonstrate the on-chip multiplexing of diseasespeci c antibodies, 1-µm-size streptavidin-coated beads were labeled with biotinylated hemagglutinin, i.e., immunodominant in uenza A-associated protein (Fig 2b; bottom) 28 . The surfaces of the 1 µm and 2.8 µm particles biofunctionalized with hemagglutinin and S-RBD, respectively, were saturated with free biotin before being added together. This successfully inhibited the aggregation of the SARS-CoV-2 and In uenza A functionalized particles (Fig. S5) and enabled size separation in the 3D nano uidic device. As a proof-of-principle experiment, the multi-particle suspension was mixed with various combinations of puri ed polyclonal rabbit antibodies, targeting either the S-RBD or hemagglutinin protein (Fig 2g). The beads were subsequently immunostained with a Cy3 anti-rabbit antibody. The obtained signal at their trapping positions is shown to correlate very well with the presence or absence of the relevant antibodies for either S-RBD (SARS-CoV-2) or hemagglutinin (In uenza A), respectively (Fig. 2h). This proof-of-principle experiment further emphasizes the versatility of our 3D nano uidic device within the framework of serological multiplexed immunoassays.
In conclusion, we have introduced a novel approach towards multiplexed antibody and disease testing by using a novel 3D PMMA-based nano uidic device. Proof of principle was obtained by showing that calibration-grade particles can be size-dependently immobilized, and their size can be accurately determined from their trapping position. The size sorting capability of the device was used to concentrate and trap S-RBD-functionalized beads in an IgG SARS-CoV-2 serological assay with a detection limit well within the range of state-of-the-art immunoassays. The test was further cross-validated with PCR-tested patient samples, showing a high degree of sensitivity and speci city. Moreover, on-bead color multiplexing has demonstrated the potential to simultaneously monitor the presence of both IgG and IgM antibodies in human sera on single particles for future time-dependent antibody studies. Additionally, we have highlighted the versatility and applicability of the 3D nano uidic device by detecting anti-S-RBD (SARS-CoV-2) and anti-hemagglutinin (In uenza A) antibodies in control samples using different bead sizes.
Here we showed multiplexing in two-dimensions, by varying particle size and the conjugated dye of the detection antibody. For future work, we aim to extend this by using various uidic channels for different immunoassays and using color-coded beads of a similar size 29 . We believe that by this multiplexing in four dimensions, concurrent detection of more than 100 antibodies on a single-chip can be achieved.
Moreover, we aim to transfer the uorescence detection to a portable and compact microscope and eventually adaptable to smartphones, since the devices do not necessitate state-of-the-art infrastructure for immunoassay read-out. We note that the facile operation and cost-effective nature of the developed 3D nano uidic device is not limited to on-chip serological immunoassays. Its applications could be further extended to various disciplines of biomedical sciences to address key research questions, ranging from mitochondrial size determination in Parkinson's disease 30 to nanoparticle-based cancer therapeutics 31 , sickle-cell diagnosis 32 and many more.

Methods
Gray scale e-beam lithography. PMMA 950K 11% in anisol was spin-coated on a 4" Si-wafer at 1000 rpm for 60 s and baked on a hotplate at 175 °C for 25 min to achieve an average lm thickness of 4 µm. The resist contrast was determined by exposing squares of 500 µm x 500 µm with e-beam doses ranging from 0 to 400 µC/cm 2 in 20 µC/cm 2 increments by using Raith EBPG 5000+ electron beam system operated at a 100 kV acceleration voltage. The PMMA 950K was developed in pure methyl isobutyl ketone (MIBK, Technic France) at 20 °C 72 h after exposure, rinsed in isopropanol (IPA, Technic France) for 20 s and blow-dried with nitrogen. The remaining resist depth of the exposed areas was measured with pro lometer (Veeco Dektak 8) using a 2.5 µm stylus and plotted against the respective exposure dose to obtain the contrast curve. The TRACER program (GenISys GmbH) was used to perform a Monte-Carlo simulation of the interaction between the incident electron beam and the resist material to obtain the point-spread function (PSF) at a given penetration depth. The nano uidic device was designed using a custom developed Python script in a GDSII format, where each layer was correlated with a different height to achieve the desired 3D topography in the trapping region and the connecting wedge. The contrast curve, PSF and nano uidic design were loaded into the BEAMER software (GenISys GmbH) to perform a 3D proximity effect correction (3D PEC) to ensure a proper dose modulation during subsequent e-beam exposure. An array of 6 devices consisting of three channels was exposed with a base dose of 80 µC/cm 2 , after which the resist was developed as previously described.
Nano uidic master fabrication and nanoimprint lithography. A 4" borosilicate wafer was cleaned with acetone and subsequently in IPA, blow-dried with nitrogen and activated with an oxygen plasma (RIE80, Oxford Instruments) at 40W with a pressure of 20 mTorr for 1 min. The adhesion promoter Ormoprime (Microresist) was spin-coated on the activated surface at 4500 rpm for 45 s and baked on a hotplate at 150 °C for 5 min. 1 mL of the UV crosslinking polymer GMN-PS90 (Optool) was pipetted directly on the surface of the developed e-beam structure, on top of which the borosilicate wafer was placed. The GMN-PS90 was allowed to spread for 20 min to ensure that all the nano uidic cavities were lled. Subsequently, the polymer was cured under UV-light at 365 nm for 6 min at a power of 300 mW/cm². As the GMN-PS90 polymer is inherently anti-sticking, the cured polymer can easily be demolded from the ebeam structures and was immediately ready for further downstream fabrication steps. More speci cally, this Optool nano uidic master was put into a Jenoptik Hex 03 imprinting tool with a 1 mm freestanding PMMA lm on top of it. To ensure a homogeneous imprinting, an anti-adhesion coated silicon wafer was placed on the backside of the PMMA lm, with a poly(amide)-PDMS-poly(amide) sandwich to equalize the imprinting pressure. A touch force of 300 N was applied and the imprinting chamber was steadily heated to a temperature of 130 °C at 9°/min before the force increased to 17 500 N, where it was kept for 15 min. The chamber was then cooled to 60 °C within 15 minutes before opening and demolding the imprinted PMMA lm from the GMN-PS90 master. The free-standing PMMA lm was activated by oxygen plasma at 80W and a pressure of 0.8 mbar for 20 s and spin-coated with a protective layer of 10% Dextran (66 kDa, Roth Industries) solution in deionized water at 3000 rpm for 60 s. The array of 6 devices was then mechanically cut into two separate 3 × 1 devices and the particle protecting Dextran layer was removed by submerging the chips in MilliQ water for 10 min and subsequent drying in a vacuum chamber at 9 mbar for 10 min.
Poly(vinyl alcohol) (PVA) functionalization and bonding. The PMMA surfaces of the 1 mm patterned lm and a 200 µm unpatterned lm were activated by UV/O-activation at 172 nm for 30 s. This reduces the molecular weight of surface polymer chains and consequently also the surface glass transition temperature (T g, surface ) with respect to the bulk material. This leads to increased interactions of PMMA chains and the surface contact area, much like in a nanovelcro-like system. Subsequently, a 0.5% PVA (Polysciences, 15132-500) solution in PBS pH 7.4 was spin-coated at 2000 rpm for 1 min. To remove any excess PVA and aid lm homogeneity, the surface of the PMMA nano uidic chip was spin-washed with deionized water at 2000 rpm for 1 min. Afterward, the patterned and unpatterned PMMA lms were aligned and bonded at 750 N and 45°C for 1 min.
Trapping of calibration grade uorescent polystyrene particles. A ve-particle mixture in PBS was made with calibration grade uorescent polystyrene particles of the following sizes and dyes: 0.9 µm (EstaporX, Bangs Laboratories Inc., FC03F-1), 1.02 µm (Yellow green, Polysciences 17154-10), 1.744 µm (Dragon Green, Polysciences 683195), 2.12 µm (Polychromatic red, Polysciences 19508-2), 3.16 µm (Yellow green, Polysciences 18861-1). A 4 µl droplet was put onto the inlet of the nano uidic device and after the liquid reached the end of the capillary pump region, a droplet of PBS was placed at the out ow to prevent premature drying of the sample. The uorescence microscopy was done with a Leica SP8 microscope using a 40x objective with 0.95 numerical aperture. A custom python script was used to analyze the average trapping positions of the particles and corresponding standard deviation. To compensate for the lateral height discrepancy in the channel, only the trapping in the central region of the channel was considered for the calculation of the trapping position. The particle sizes speci ed by the supplier were used to t a linear regression curve to the particle trapping position. The mean particle size and coe cient of variation were recalculated from the linear regression t. with three subsequent washing steps. To control the ow and trapping functionality of the device, 0.9 µm non-magnetic Estaporix-loaded polystyrene beads were added to the suspension. A 4µl droplet of the suspension was then loaded onto the 3D nano uidic PMMA chip and after the liquid reached the end of the capillary pump, a droplet of PBS (pH 7.4) was pipetted on the out ow region to halt the further uid ow. The uorescent signal in the relevant trapping region was quanti ed by using a custom Python image analysis script. The script used the scikit-image module to identify the relevant uorescent pixels, perform the background subtraction and the equalization with a top-hat lter. Afterward, NumPy and Pandas were used to calculate the mean signal of the acquired image. Assay Validation. The assay described above was performed with a 29-patient sample set (Anawa GmbH, COV-POSSET-S), consisting of 19 IgG positive and 10 IgG negative patients. All of the IgG-positive patients had a positive PCR test for COVID-19, whereas the IgG-negative patients had a negative PCR test. The highest uorescence signal of the negatively-tested patient population was taken as a cut-off value at 6.159 arbitrary units. A higher signal was considered to be a positive result.
Antibody and disease multiplexing. Antibody multiplexing. Dynabeads M-280 were labeled with the COVID-19 spike RBD as previously outlined. Afterward, they were mixed in a 3:4 ratio and incubated with serum from a positive patient control with a high ELISA titer for IgG and IgM antibodies against the spike RBD as well as with a negative patient control, which was tested PCR negative for COVID-19. To achieve color-multiplexed antibody detection, two secondary antibodies were added to the particle suspension: a donkey anti-human IgM Alexa 488 (Bioconcept, 2020-30) and donkey anti-human IgG Cy5 (Jackson Immuno, AB_2340539). Both were incubated with the washed particle suspension as described in the previous section. The device loading and investigation as well as associated analysis, were performed in the same fashion. Disease multiplexing. As a proof-of-principle, a particle suspension containing spike RBD functionalized Dynabeads M-280 and hemagglutinin functionalized Dynabeads MyOne T1 particles were mixed with puri ed polyclonal rabbit anti-spike (Sino Biological, 40592-T62) and puri ed polyclonal rabbit anti-hemagglutinin (Sino biological, 11055-RP01). The beads were then immunostained with donkey anti-rabbit Cy3 (Jackson Immuno, AB_2307443) for 1 h at 25°C and 600 rpm after which the suspension was washed three times with PBS (pH 7.4). 0.9 µm Estaporix ow control beads were added before loading the particles into the nano uidic device. The relevant trapping lines were investigated and analyzed as previously described.