Contents lists available at ScienceDirect

Micro and Nano Engineering

journal homepage: www.journals.elsevier.com/micro-and-nano-engineering

Research paper

Biosensing based on optimized asymmetric optofluidic nanochannel gratings

Foelke Purra,b, Rachel D. Loweb, Matthias Stehrc, Mahavir Singhc, Thomas P. Burgb,d,Andreas Dietzela,⁎

a Technische Universität Braunschweig, Institute of Microtechnology, Braunschweig 38124, GermanybMax Planck Institute for Biophysical Chemistry, Göttingen 37077, Germanyc Lionex Diagnostics & Therapeutics GmbH, Braunschweig 38124, Germanyd Technische Universität Darmstadt, Department of Electrical Engineering and Information Technology, Darmstadt 64283, Germany

A R T I C L E I N F O

Keywords:InterferometryPoint-of-careNon-diffusion-limitation

A B S T R A C T

Interferometric biosensing offers the potential of label-free detection with very high sensitivities. In our opto-fluidic grating detector the accumulation of biological target molecules on the surface of functionalized na-nochannels can be detected due to the change in refractive index. Based on an analytical model, the sensitivity ofthe optofluidic grating detector was improved by a factor of 3.5 to 8.45 RIU−1 (refractive index unit) byadapting the grating geometries. The grating consists of nanochannels, through which the sample flows. Thisallows transport without diffusion-limitations inside the sensing area, as it is of particular importance whenmolecule concentration in the sample decreases. The nanochannels were functionalized by silanization andcovalent bonding of antibodies. Finally, the binding of Mycobacterium ulcerans specific antigen MUL2232 wasdetected label-free.

1. Introduction

Miniaturization of analytical instruments enables the potential toincrease performance while simultaneously reducing costs. Microfluidicplatforms, in particular, require very small volumes which drasticallyreduces sample consumption and costs, while also significantly short-ening the time from sampling to results. All these features, thereforemake microfluidic systems the ideal platform for point-of-care tests [1].

Here, optofluidics, as a combination of microfluidics and opticalanalysis methods, is of special interest. With methods such as surfaceplasmon resonance SPR [2–4], photonic crystals [5] and others [6],sensitive and label-free biosensors have already been realized.

All of the above label-free biosensors rely on surface-based sensing.As such, when the target molecule concentration in the solution de-creases, analyte depletion in combination with diffusion-limited trans-port to the sensor surface hinder the detection process [7]. When de-signing a microfluidic biosensor, this aspect needs then to be consideredto ensure that a sufficient numbers of target molecules interact with thedetector in a short enough time.

In our previously introduced asymmetric nanofluidic grating de-tector the measurement is performed in the Fourier space. Fluidicchannels are arranged as an optical grating and the refractive indexdifference between reference and detection channels is determined

optically. A sensitivity of 2.416 RIU−1 (refractive index units) and thedetection of an adsorbed protein layer in the detection channels hasbeen successfully achieved [8].

Here we show how the sensitivity of our asymmetric nanofluidicgrating detector can be improved significantly by adapting the geo-metrical arrangement based on an analytical model.

The optical biosensor is functionalized and detects specific bindingof target molecules as small as 18 kDa without further labeling. Masstransport can be then shown to be dominated by convection throughthe channel rather than by diffusion across a boundary layer. In thisregime, all molecules passing through the nanochannel device en-counter the surface with high probability, making them available toparticipate in a binding reaction.

1.1. Asymmetric nanofluidic grating biosensor

In Fig. 1 a schematic of the optical grating made of nanochannels isshown. Reference and detection channels are placed in an inter-digitating structure. Collimated laser light is diffracted by the opticalgrating and the resulting diffraction pattern is analyzed. The intensitydistribution will change sensitively with a change in refractive indexdue to protein accumulation at the channel walls.

Vias at the end of each nanochannel connect with the bigger supply

https://doi.org/10.1016/j.mne.2020.100056Received 16 April 2020; Accepted 2 May 2020

⁎ Corresponding author.E-mail address: a.dietzel@tu-braunschweig.de (A. Dietzel).

Micro and Nano Engineering 8 (2020) 100056

2590-0072/ © 2020 Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).

T

channels. Through these, detection and reference structures can beflushed with different fluids from separate supply channels or rinsed bya common supply channel.

Importantly, the unit cell of the grating shows an asymmetricstructure (see Fig. 2D and SI 1). The interdigitating channels of thereference structure are shifted out of the middle position between thedetection channels. This arrangement allows the direct sensitive de-tection of refractive index differences between the two sets of channelswhich provides an inherent common mode rejection for small refractiveindex differences, as reported earlier [8].

2. Fabrication

The asymmetric nanofluidic chip is made of silicon and glass al-lowing the treatment with harsh chemicals like piranha solution aftereach experiment. Therefore, chips can be reused multiple times.Additionally, the surface chemistry of silicon and glass is very similar,allowing the same functionalization chemistry for later immunoassaydevelopment and antibody immobilization.

The fabrication protocol has been described before [8,9]. Fig. 2shows SEM images of the final grating made of nanochannels. Thegrating consists of 25 reference and detection channels. Each channelhas a width of w = 3 μm per channel and a channel depth ofh = 290 nm. The asymmetry of the unit cell can be described by theasymmetry parameter l, here l = 9.5 μm (Fig. 2D).

3. Optical and fluidic setup

The chip is placed inside an optical setup, as shown schematically inFig. 3. Here, a collimated laser beam (wavelength λ = 635 nm) is firstfocused to reduce the beam diameter to approximately 350 μm andpasses through a hole in a parabolic mirror before it is perpendicularly

incident on the fluidic grating surface. The diffracted laser light is re-flected back towards the parabolic mirror, which reflects and focusesthe diffraction pattern towards a sCMOS camera (PCO edge 4.2, Kel-heim, Germany). Due to the aperture of the mirror, the 0th maximum isnot recorded by the camera. Higher order maxima (m ≠ 0) are there-fore not overwhelmed by the center maximum, as this is very intensebut contains no information for the measurement. Before the laser hitsthe diffraction grating, the intensity is reduced by a grey filter. In thisway the intensity of the diffraction maxima with m ≠ 0 does notoversaturate the camera sensor. However, the intensity is still highenough to set the exposure time as short as possible and to record moreimages per time. Using this procedure and subsequent averaging of theimages, impact of white noise sources like shot noise, is reduced [10].

A region of interest (ROI) is set around two maxima of the sameorder and the intensity Im of each is determined by integrating the greyvalues. Finally, the signal is calculated from Eq. (1).

Fig. 1. Schematic of the nanofluidic grating detector. A) Nanofluidic channelsare arranged as an optical grating, where reference and detection channels areplaced in an interdigitating structure. Each channel is connected to biggersupply channels. B) Laser light is diffracted by the optofluidic grating. Everysecond channel is a detection channel, which is functionalized with specificantibodies. Target molecules accumulating inside the detection channels willalter the intensity distribution of the diffraction pattern due to a local change inthe refractive index.

Fig. 2. SEM images of chip structures A) optical grating made of nanochannels,B) Cross sections B-B of one via and a parallel nanochannel, C) Cross section A-A: The nanochannels are created by wet chemical etching of a thin silicon oxidelayer on a silicon wafer and subsequently sealed with a glass wafer on top byanodic bonding. D) Cross-section showing detail of grating cross section withperiod P, asymmetry l and channel width w.

Fig. 3. Schematic of the optical setup.

F. Purr, et al. Micro and Nano Engineering 8 (2020) 100056

2

= −+

−

−S I I

I Imm m

m m (1)

The chip is mounted inside a fluidic holder and pressure controllersensure the fluidic flow. Here, the separate supply channels for referenceand detection channels allow the specific functionalization of the de-tection channels, without affecting the reference structures. However,both types of nanochannels can be rinsed from the third supply channelsimultaneously (see SI 2). The applied pressure values are presented inTable 1 of SI 2.

4. Analytical modelling and experimental characterization

The refractive index n can be used as a measure of sample con-centration [6]. The refractive index influences the optical phase shift ϕinside the nanochannels of the optofluidic grating, which is defined as

=φ πλ

hn22(2)

Consequently, the difference in the optical phase Δϕ between re-ference and detection channels has a direct impact on the intensitydistribution of the diffraction pattern generated by the optofluidicgrating. To evaluate the dependency of refractive index difference onthe optical signal, an analytical model was established.

The intensity distribution of the diffraction pattern can be analyti-cally calculated by a Fourier transformation of the grating [11].

The obtained signal (Eq. (3)) depends on the refractive index dif-ference Δn between reference and detection channels and therewith thephase difference Δϕ, and the order of maximum m. This is valid forsmall Δϕ as we expect it for protein binding [8]. As a third parameter,the asymmetry l/P has a certain impact on the signal. Here, l/P de-scribes how much the unit cell structure is shifted from the symmetricarrangement, with the two extreme cases l/P = .5 for a symmetricgrating and l/P = 1-w, when reference and detection channels are indirect contact to each other.

= −+

≈ − ⎛⎝

⎞⎠

∆ + ∆−

−S I I

I Itan l

Pπm φ φ1

2( )m

m m

m m

3O(3)

We analyzed the influence of asymmetry (l/P) on the sensitivity ofour biosensor on refractive index changes by analytical calculation. Theresults for the first two order of maxima are shown in Fig. 4. The in-tensity difference between positive and negative maxima of the sameorder ΔIm shows a strong dependency on the phase shift Δϕ. However,also the asymmetry parameter l/P has a certain impact on how sensitivethe signal changes due to a change in phase.

As small changes in refractive index due to protein binding are

expected, the sensitivity values ∆ ≈∆ φ( 0)dSd n are of special interest.

Here, the first order shows clear optima for l/P = .507 and 0.542, thesecond order for l/P = .727. For the second order, an additional op-timum appears around l/P = .772. However, this is challenging tofabricate, as the distance between detection and reference channelsneeds to be very small. For instance, in a grating with a period of 16 μmthe reference and detection channel would need to be separated by only200 nm on one side.

We built two types of gratings with different asymmetry (l/P = .59and 0.72) and evaluated the response on varying refractive index dif-ferences. For this purpose, the detection channels were filled withglycerol solutions of different concentrations (2%–8% w/w in water)representing a range for refractive index values of n = 1.3354–1.3434as measured with a handheld refractometer (DR201–95, A.Kruess,Hamburg, Germany). The reference was filled with water (n = 1.3325)for the entire experiment. Each experiment was repeated at least threetimes and the results are plotted in Fig. 5.

The results of the experiments are in good agreement with theanalytical findings. For an asymmetric parameter l/P = .59 the slope ofthe signal for the first order (sensitivity = 8.45 RIU−1) is much highercompared to the second order. However, the sensitivity of the secondorder can be improved by using a device with an asymmetry parameterof l/P = .72. Here, the sensitivity was increased to −11.62 RIU−1.

Fig. 4. Influence of Δφ and l/P on ∆ = −+ −Im I Im m and the sensitivity∆

dSmd n

(h = 328 nm, P = 16 µm).

Fig. 5. Signal dependency on refractive index differences for the first andsecond order with asymmetry parameters l/P = .59 and 0.72. The results of theanalytical model (lines) are validated by experiments (marks).

F. Purr, et al. Micro and Nano Engineering 8 (2020) 100056

3

Fabrication of the l/P = .72 chips requires a lithographic resolutionbetter than 2 μm, which is challenging with a standard contact align-ment process. Therefore the fabrication yield was relatively low and wedecided to continue the protein experiments with the l/P = .59 chipsevaluating the first order maxima. With a sensitivity of 8.45 RIU−1 westill gain a 3.5-fold higher sensitivity compared to the first generation ofchips [8].

5. Fluidic flow and molecule diffusion inside the nanograting

For biosensing, the channel surfaces need to be functionalized withspecific capture molecules and the accumulation of the respective targetmolecules is measured (see Section 6). In addition to the optical de-tection principle, it is also important to understand the interaction ofcapture and target molecules in the detection area, which is caused bythe transport and reaction conditions within the fluidic channels.

Therefore, the velocity inside the nanochannels was determined byfluorescence measurements. The grating was placed under a fluores-cence microscope and the supply channel of the detection channels wasfilled with fluorescein sodium salt in water. The propagation of thefluorescence dye in the grating was recorded with a camera (PhantomVEO 640 L, Vision Research, USA) after changing the fluidic pressureconfiguration from rinsing to functionalization (see Table 1 of SI2).

By evaluating the distance of the fluorescence front over time, wecalculated an average velocity of about 0.0002 m/s (Fig. 6).

The Dammköhler number Da is a dimensionless number to describethe relation between rate of arrival at the sensing surface by diffusionaltransport for the target molecules and the rate of reaction with thecapture molecules. Da ≫ 1 indicates a diffusion-limited process andDa ≪ 1 indicates that the process becomes reaction-limited [12,13].

= ∙ ∙Da k C hD

on s0(4)

with Cs0 – total number of binding sites kon – association rate constant h– channel height D – diffusivity

As an example, we consider the binding process of a secondary IgGantibody to the related IgG antibody on a functionalized surface, whichreveals Da = 0.0018 (assuming Cs0 = 10.13 fmol/mm2 [14];kon = 2.73 × 104 M−1 s−1 [15]; D = 4.4 × 10−5 mm2s [16] andh = 290 nm), indicating that the binding between two antibodies isreaction-limited inside the sensing grating.

As illustrated by Fig. 7, this will lead to rapid axial depletion oftarget molecules. Thereby it is ensured that every molecule will arriveat the channels surface activated with capture molecules.

However, as the fluidic flow is slow due to the small structure sizeand the low pressure difference applied, the reaction process over theentire channel length becomes convection limited. This is characterizedby the reaction/convection number ζDa [12]

= ∙ ∙∙

ζDa k C LU h

on s0(5)

with L – channel length U – velocityIn this example ζDa = 1.5, indicating convection-limited conditions

(ζDa > 1) [12]. Binding occurs right at the entrance of the grating anddepletes the bulk solution. As the surface saturates, the depletion frontmoves through the nanochannel in axial direction in a wave like fashionleaving behind saturated surfaces.

6. Immunoassay for antigen detection

For the application as biosensor, the reliable and stable im-mobilization with capture molecules like antibodies is essential.Consequently, we used a covalent coupling process of specific anti-bodies to the nanochannel walls in order to enable the specific accu-mulation of target molecules only in the detection channels.

First, all nanochannels are silanized with APTMS (3-aminopropyltrimethoxysilane) in the gas phase. After washing with ethanol, waterand PBS (sodium phosphate buffer), glutaraldehyde is introduced onlyto the detection channels. Glutaraldehyde binds to the amine group ofthe silane and offers a reactive aldehyde group to bind to an aminegroup of the antibody. [17–20]. The functionalization process is illu-strated in Fig. 8.

Due to the accumulation of proteins inside the fluidic channels, theoptical path length is increased. As proteins exhibit a higher refractiveindex in the range of 1.36–1.55 [21] compared to the buffer medium(e.g. refractive index of PBS = 1.334) this will lead to a change in theaverage refractive index within the detection channels compared to thereference channels, which are filled with buffer medium only. Thisrefractive index difference can be detected in the same manner as in themeasurements with glycerol solutions as shown in Fig. 5.

6.1. Proof of antibody coupling and detection

The functionalization process described above has been used to

Fig. 6. Velocity determination by running fluorescent dye through the detection channels.

Fig. 7. Schematic of the depletion of target molecules inside a nanofluidicchannel

F. Purr, et al. Micro and Nano Engineering 8 (2020) 100056

4

immobilize anti-mouse IgG antibodies produced in goat on the detec-tion channel surfaces. This increases the optical path difference be-tween the detection channels and the reference channels. In the fol-lowing, we display the resulting path differences in terms of anequivalent refractive index difference Δn to facilitate comparison withFig. 5. Fig. 9 gives the equivalent refractive index difference over timeduring the functionalization and subsequent detection of the respectiveanti-goat antibody. The accumulation of the antibody leads to an in-crease in the effective refractive index in the detection channels. As thereference channels are not affected, Δn increases due to the formationof the antibody layer (reference channels remain filled with only PBS).This shift is stable even after rinsing the complete system with PBS. Theimmobilization of the antibodies was additionally proven by introdu-cing anti-goat IgG through the detection channels. Again, the signalincreases, indicating the binding of the secondary antibody on thefunctionalized channel walls.

We repeated the experiment with different anti-goat IgG con-centrations. As the surface coverage varied from experiment to ex-periment, due to functionalization reproducibility, the signal changewas scaled by the maximum signal after functionalization Δnmax, func

with the primary antibody. The results are shown in Fig. 10. For theexperiments with sencondary antibody concentrations of 36 μg/ml and120 μg/ml the fluctuations are stronger compared to the experimentwith 72 μg/ml. Variations in the surface coverage along the channeldue to imperfect cleaning or partial blockage of individual

nanochannels during the measurement might at least partially explainthe varying strength of the fluctuations. However, this will need to beinvestigated in more detail in the future. Nevertheless, there is a clearrelationship between the signal slope and the anti-goat IgG concentra-tion. Also, the equilibrium signal tends towards higher levels for higherconcentrations of secondary antibody (see SI 3).

6.2. Detection of the Mycobacterium ulcerans antigen MUL2232

Buruli ulcer is an infectious disease of the skin and soft tissues,which is particularly prevalent in the tropics. After tuberculosis andleprosy, it is the most common disease worldwide caused by myco-bacteria [22]. A major challenge to prevent the disease is the lack ofappropriate diagnostic procedures since PCR based tests are not avail-able in most remote areas. Hence, the diagnosis of buruli ulcer in theaffected areas is usually made much too late, when a clear clinicalpicture already appears. This calls for a fast, sensitive and reliablepoint-of-care test.

The development of a specific biosensor for Mycobacterium ulceransis complicated due to the antigenic cross-reaction among differentmycobacterium species. Recently, small heat shock proteins expressed bythe bacterium and located on the cell wall surface, have attractedspecial attention. These proteins are highly specific and enable theantigen capture-based detection [23].

Fig. 11 shows the signal for functionalization and detection of theantigen MUL2232, which is a small heat shock protein ofMycobacteriumulcerans with a molecular weight of 18 kDa [24].

First, the complete grating is rinsed with PBS, while the referencechannels are filled with PBS for the entire experiment. Glutaraldehyde(10% in PBS) is introduced into the detection channels. Glutaraldehydehas a higher refractive index compared to PBS indicated by the highshift in the signal curve. Afterwards the antibody (anti-MUL2232,0.5 mg/ml in PBS) is pushed through the detection channels. The signalchanges from 0.003 to 0.0055 due to the binding on the channel walls.Even after washing with PBS, the signal stays on the higher level, in-dicating the formation of a stable protein layer. Finally, the buruli an-tigen MUL2232 (0.5 mg/ml) is introduced. The increase of the signalindicates the binding of MUL2232 to the antibody.

This shows that we are able to functionalize nanochannels of anoptofluidic grating with antibodies and to detect binding without anyfurther labeling of the target molecule, despite the fact that the mole-cular weight of MUL2232 is only 18 kDa [24].

Fig. 8. Functionalization of the detection channels by covalent bonding of specific antibodies with a silane coupling. For details see text. APTMS (3-aminopropyltrimethoxysilane). OHC(CH2)3CHO (Glutaraldehyde). The antibody is shown in yellow. (For interpretation of the references to colour in this figure legend, thereader is referred to the web version of this article.)

Fig. 9. Measured refractive index difference for the functionalization with anti-mouse antibodies (1.15 mg/ml in PBS) and specific binding of anti-goat anti-bodies (72 μg/ml in PBS + 1 mg/ml BSA).

F. Purr, et al. Micro and Nano Engineering 8 (2020) 100056

5

7. Conclusion and outlook

For the detection of specific biomolecules a high sensitivity and athorough understanding of the transport phenomena is essential togenerate reliable signals for both small analytes and low concentra-tions.

Here, we present our optofluidic biosensor based on refractive indexdifference measurements in the Fourier space. We were able to improvethe sensitivity of our sensor over 3.5 fold by adapting the grating designbased on an analytical model. The results were then validated withexperimental trials. Further improvements might be feasible by otherfabrication technologies like e-beam lithography or focus ion beamstructuring to achieve more accurate distances between the na-nochannels.

Specific biomolecule binding was detected after functionalization ofthe detection channels with antibodies by covalent binding via silani-zation. In this way, we were able to detect the binding ofMycobacteriumulcerans antigen MUL2232. However, the samples were prepared infiltered buffer medium with concentrations above the expected valuesfound in field (lysate concentration in the range of a few ng/ml [23]).The significance of these results, is that the detection of proteins with amolecular weight of only 18 kDa can be realized without any sub-sequent labelling. This shows great promise for future studies with fieldsamples.

With channel heights below 300 nm, mass transport to the sensorsurface is limited by convection rather than by diffusion. This allows a

complete depletion of the target molecules inside the nanochannels.This regime is advantageous for precise and fast point-of-care diag-nostic by end-point detection, as it can be ensured that every targetmolecule reacts with the functionalized surface in a short time.

Bridging the gap between highly specialized research and the ap-plication in the “real-world” as an affordable and location-independent,but sensitive POC device remains a major challenge of many micro-fluidic biosensors to still be addressed [25,26].

In a previous study, we were able to translate the laboratory-basedsetup to a mobile POC setup including a smartphone-assisted readoutdevice [9]. Low sample volumes with no diffusion limitations, easyhandling by technically untrained persons, no need for sophisticatedequipment and the inherent robustness against biasing signals liketemperature fluctuations, make the asymmetric nanofluidic gratingdetector a promising tool for POC applications. However, the fabrica-tion based on silicon and glass processing to create nano and microscale fluidic structures currently does not translate well into low-costlarge scale production [26]. Even though the fabrication process basedon silicon and glass allows the creation of smallest feature sizes withhigh precision and tailor-made 3D fluidic connections like the 3 μm viasat the end of each nanochannel, it is still labor and cost intensive andnot standardized leading to higher chip-to-chip variations [26]. For theapplication as an affordable and disposable POC device, low cost al-ternatives are required. Here, fabrication based on polymer materialsoffers the potential of cost effective, large-scale production. Some au-thors already have shown that even complicated 3D and nanofluidic

Fig. 10. Measured refractive index (normalized) for different anti-goat concentrations

Fig. 11. Covalent binding of antibodies to the channels walls and detection of buruli ulcer specific antigen MUL2232.

F. Purr, et al. Micro and Nano Engineering 8 (2020) 100056

6

structures can be realized based on PDMS [27], SU-8 [28], polymethylmethacrylate (PMMA) [29] or cycloolefin polymer (COP) [30].

Funding

This project was funded within the program “ZentralesInnovationsprogramm Mittelstand” by the Bundesministerium fürWirtschaft und Energie (BMWi) (KF3259501MD3/KF3085602MD3).Further funding was provided by a Max Planck Research Group award(T. P. Burg) and by the Max Planck Institute for Biophysical Chemistry.

Declaration of Competing Interest

None.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.mne.2020.100056.

References

[1] R.C. Dunn, Compact, inexpensive refractive index detection in femtoliter volumesusing commercial optical pickup technology, Anal. Methods 11 (2019) 2303–2310,https://doi.org/10.1039/c9ay00369j.

[2] H.H. Nguyen, J. Park, S. Kang, M. Kim, Surface plasmon resonance: a versatiletechnique for biosensor applications, Sensors 15 (2015) 10481–10510, https://doi.org/10.3390/s150510481.

[3] Y. Qian, X. Zeng, Y. Gao, H. Li, S. Kumar, Q. Gan, X. Cheng, F.J. Bartoli, Intensity-modulated nanoplasmonic interferometric sensor for MMP-9 detection, Lab Chip 19(2019) 1267–1276, https://doi.org/10.1039/c8lc01391h.

[4] A.E. Cetin, A.F. Coskun, B.C. Galarreta, M. Huang, D. Herman, A. Ozcan, H. Altug,Handheld high-throughput plasmonic biosensor using computational on-chip ima-ging, Light Sci. Appl. 3 (2014) e122, https://doi.org/10.1038/lsa.2014.3.

[5] I. Petrova, V. Konopsky, I. Nabiev, A. Sukhanova, Label-free flow multiplex bio-sensing via photonic crystal surface mode detection, Sci. Rep. 9 (2019) 8745,https://doi.org/10.1038/s41598-019-45166-3.

[6] X. Fan, I.M. White, S.I. Shopova, H. Zhu, J.D. Suter, Y. Sun, Sensitive optical bio-sensors for unlabeled targets: a review, Anal. Chim. Acta 620 (2008) 8–26, https://doi.org/10.1016/j.aca.2008.05.022.

[7] D.S. Jin, E.L. Brightbill, E.M. Vogel, General model for mass transport to planar andnanowire biosensor surfaces, J. Appl. Phys. 125 (2019) 114502, https://doi.org/10.1063/1.5084253.

[8] F. Purr, M. Bassu, R.D. Lowe, B. Thürmann, A. Dietzel, T.P. Burg, Asymmetric na-nofluidic grating detector for differential refractive index measurement and bio-sensing, Lab Chip 17 (2017) 4265–4272, https://doi.org/10.1039/c7lc00929a.

[9] F. Purr, M.-F. Eckardt, J. Kieserling, P.-L. Gronwald, T.P. Burg, A. Dietzel, Robustsmartphone assisted biosensing based on asymmetric nanofluidic grating inter-ferometry, Sensors 19 (2019), https://doi.org/10.3390/s19092065.

[10] D.B. Murphy, M.W. Davidson, Fundamentals of Light Microscopy and ElectronicImaging, 2nd ed., Wiley-Blackwell, Hoboken (Nueva Jersey), 2013.

[11] F. Pedrotti, Optik für Ingenieure: Grundlagen, 3rd edn, Springer Verlag, Berlin,2005.

[12] T. Gervais, K.F. Jensen, Mass transport and surface reactions in microfluidic sys-tems, Chem. Eng. Sci. 61 (2006) 1102–1121, https://doi.org/10.1016/j.ces.2005.06.024.

[13] T.M. Squires, R.J. Messinger, S.R. Manalis, Making it stick, Nat. Biotechnol. 26(2008) 417–426, https://doi.org/10.1038/nbt1388.

[14] S.K. Vashist, C.K. Dixit, B.D. MacCraith, R. O’Kennedy, Effect of antibody im-mobilization strategies on the analytical performance of a surface plasmon re-sonance-based immunoassay, Analyst 136 (2011) 4431–4436, https://doi.org/10.1039/c1an15325k.

[15] L.L. Yin, S.P. Wang, X.N. Shan, S.T. Zhang, N.J. Tao, Quantification of protein in-teraction kinetics in a micro droplet, Rev. Sci. Instrum. 86 (2015) 114101, https://doi.org/10.1063/1.4934802.

[16] W.M. Saltzman, M.L. Radomsky, K.J. Whaley, R.A. Cone, Antibody diffusion inhuman cervical mucus, Biophys. J. 66 (1994) 508–515, https://doi.org/10.1016/s0006-3495(94)80802-1.

[17] G.T. Hermanson, Bioconjugate techniques, Academic Press, Amsterdam, 2008.[18] Y. Wang, R.D. Lowe, Y.X. Mejia, H. Feindt, S. Steltenkamp, T.P. Burg, A fluores-

cence based method for the quantification of surface functional groups in closedmicro- and nanofluidic channels, Biomicrofluidics 7 (2013) 26503, https://doi.org/10.1063/1.4802270.

[19] M. Kim, S.O. Jung, K. Park, E.-J. Jeong, H.-A. Joung, T.-H. Kim, D.-W. Seol,B.H. Chung, Detection of Bax protein conformational change using a surfaceplasmon resonance imaging-based antibody chip, Biochem. Biophys. Res. Commun.338 (2005) 1834–1838, https://doi.org/10.1016/j.bbrc.2005.10.155.

[20] D. Kim, A.E. Herr, Protein immobilization techniques for microfluidic assays,Biomicrofluidics 7 (2013) 41501, https://doi.org/10.1063/1.4816934.

[21] J. Vörös, The density and refractive index of adsorbing protein layers, Biophys. J.87 (2004) 553–561, https://doi.org/10.1529/biophysj.103.030072.

[22] E. Weir, Buruli ulcer, Can. Med. Assoc. J. 166 (2002) 1691.[23] A. Dreyer, K. Röltgen, J.P. Dangy, M.T. Ruf, N. Scherr, M. Bolz, N.J. Tobias,

C. Moes, A. Vettiger, T.P. Stinear, G. Pluschke, Identification of the Mycobacteriumulcerans protein MUL_3720 as a promising target for the development of a diag-nostic test for Buruli ulcer, PLoS Negl. Trop. Dis. 9 (2015) e0003477, , https://doi.org/10.1371/journal.pntd.0003477.

[24] M. Bolz, A. Bénard, A.M. Dreyer, S. Kerber, A. Vettiger, W. Oehlmann, M. Singh,M.S. Duthie, G. Pluschke, Vaccination with the surface proteins MUL_2232 andMUL_3720 of Mycobacterium ulcerans induces antibodies but fails to provide pro-tection against Buruli ulcer, PLoS Negl. Trop. Dis. 10 (2016) e0004431, , https://doi.org/10.1371/journal.pntd.0004431.

[25] M.S. Chiriacò, M. Bianco, A. Nigro, E. Primiceri, F. Ferrara, A. Romano,A. Quattrini, R. Furlan, V. Arima, G. Maruccio, Lab-on-chip for exosomes and mi-crovesicles detection and characterization, Sensors 18 (2018), https://doi.org/10.3390/s18103175.

[26] M.I. Mohammed, S. Haswell, I. Gibson, Lab-on-a-chip or Chip-in-a-lab, Proc.Technol. 20 (2015) 54–59, https://doi.org/10.1016/j.protcy.2015.07.010.

[27] D. Pezzuoli, E. Angeli, D. Repetto, P. Guida, G. Firpo, L. Repetto, Increased flex-ibility in lab-on-chip design with a polymer patchwork approach, Nanomaterials 9(2019), https://doi.org/10.3390/nano9121678.

[28] M.S. Chiriacò, M. Bianco, F. Amato, E. Primiceri, F. Ferrara, V. Arima, G. Maruccio,Fabrication of interconnected multilevel channels in a monolithic SU-8 structureusing a LOR sacrificial layer, Microelectron. Eng. 164 (2016) 30–35, https://doi.org/10.1016/j.mee.2016.07.006.

[29] Z. Yin, L. Sun, H. Zou, E. Cheng, Two dimensional PMMA nanofluidic device fab-ricated by hot embossing and oxygen plasma assisted thermal bonding methods,Nanotechnology 26 (2015) 215302, https://doi.org/10.1088/0957-4484/26/21/215302.

[30] L. Yi, W. Xiaodong, Y. Fan, Microfluidic chip made of COP (cyclo-olefin polymer)and comparion to PMMA (polymethylmethacrylate) microfluidic chip, J. Mater.Process. Technol. 208 (2008) 63–69, https://doi.org/10.1016/j.jmatprotec.2007.12.146.

F. Purr, et al. Micro and Nano Engineering 8 (2020) 100056

7