Fiber optic interferometric immunosensor based on polydimethilsiloxane (PDMS) and bioactive lipids

: We demonstrate a novel and simple means to fabricate optical ﬁber immunosensors based on Fabry-Perot (F-P) interferometers using polydimethylsiloxane (PDMS) as support for bioactive lipids. The sensors are fabricated following a straightforward dip-coating method producing PDMS end-capped devices. A biosensing platform is realized by subsequent functionalization of the PDMS cap with a previously characterized bioactive lipid antigen cocktail from Mycobacterium fortuitum , used as a surrogate source of antigens for tuberculosis diagnosis. After functionalization of the PDMS, the F-P sensors were immersed in diﬀerent antibody-containing sera and the registered changes in their spectral features were associated to the interactions between the active lipids and the serum antibodies. Our results show that the proposed PDMS end-capped F-P immunosensors perform well diﬀerentiating antibody-containing sera. Furthermore, they oﬀer attractive attributes such as label-free operation, real-time detection capabilities and they are also reusable. The proposed sensors, therefore, serve as an enabling optical immunosensing technique oﬀering excellent potential for developing novel lipidomic analytical tools.


Introduction
Fiber-optic Fabry-Pérot interferometers (FFPIs) have been used extensively for measuring chemical and physical parameters. Strain, molar concentration, temperature and refractive index, to name a few, have been successfully monitored using fiber-based devices [1]. Amid the different techniques for fabricating extrinsic Fabry-Perot (FP) cavities, the use of polymer overlays on the tip of standard single-mode optical fibers has gained a lot of attention [2][3][4][5][6][7]. This is mostly due to the ease of processing of polymeric materials, which can be deposited following straightforward dip-coating procedures, and can be readily cured on the tip of the fibers. Several polymers have been shown to provide a suitable platform for developing the so-called polymer end-caps forming the F-P cavity in these structures. Refractive index, temperature, pressure, humidity and concentration measurements have been carried out with such configurations [2][3][4][5][6]8], yielding robust structures and in some cases with multi-variable sensing capabilities (e.g., pressure and temperature) [7]. For biosensing applications, adequate biocompatible platforms capable to support bioactive molecules for biorecognition are always required. Polymer platforms are commonly used for these purposes, and they typically require further treatment to promote its functionalization with these bioactive entities [9][10][11].
Among the available polymers, Polydimethylsiloxane (PDMS) has become one of the main choices for biomedical related applications owing to its biocompatibility, chemical inertness, haemocompatibility, and optical transparency [12][13][14][15]. It is further easy to process, flexible and has low surface energy allowing for creating complex arrays for microfluidic biosensors and for miniaturizable detection systems [16][17][18]. However, the high hydrophobicity of PDMS [19] has been considered as a disadvantageous feature for biosensing applications because these are mostly performed on aqueous-based systems. Biosensors generally rely on mediators of biological activity, including enzymes and other proteins, nucleic acids or peptides; consequently, several methods have been envisioned for surface modification of PDMS in order to promote the attachment of these hydrophilic molecules [15,[20][21][22][23]. In contrast, the use of this versatile polymer for the attachment of molecular entities with total or partial hydrophobicity, such as lipidic molecules, has not yet been explored. In this paper, we demonstrate the use of PDMS functionalized with bioactive lipids as a suitable material for the development of a end-capped FFPI immunosensor.
The functionalization of the PDMS with bioactive lipids was performed by physisortion, via drop-casting, based on the hydrophobic nature of both, the polymer and the lipid [24]. To evaluate the performance of the proposed immunosensor, we used a lipid antigen cocktail from Mycobacterium fortuitum and the antibodies capable to recognize these lipid antigens. Such bacterial species have been previously used as a surrogate source of antigens for tuberculosis diagnosis [25,26]. The reactivity and differentiation capabilities of antibody-containing sera of the lipid-functionalized PDMS immunosensor was evaluated through spectral analysis. In particular, changes in the surrounding refractive index and dimensions of the end-cap, associated to the interactions between the active lipids and the serum antibodies, were effectively monitored by analyzing the spectral features of the FFPI. Our results demonstrate that the immunosensor is indeed able to recognize the sera antibodies offering an appropriate means for sera differentiation. We thus demonstrate an attractive lipidic platform for developing label-free immunosensors with real-time and direct detection capabilities.

Materials and biological samples
A Sylgard 184 silicone elastomer (PDMS) kit was purchased from Dow Corning, USA. Phosphate buffer saline (PBS) 1M was prepared by adding 80 g of NaCl, 2 g of KCl, 14.4 g of Na2HPO4 and 2.4 g of KH2PO4 to 800 mL of deionized water, adjusting to pH=7.4 and diluted to the 1 L mark of a test tube. 3% solution of BSA (purchased from Sigma Aldrich, USA) in PBS was used as blocking buffer. A denaturing potassium biphthalate buffer was used to detach antibodies from the antigen-functionalized PDMS. This buffering solution was prepared by mixing 100 ml 0.1 M potassium hydrogen phthalate and 44.6 ml of 0.1 M HCl, followed by pH adjustment to 3.0.
The bacterial lipid antigens for PDMS functionalization were extracted from Mycobacterium fortuitum, comprising a mixture of glycolipids of high antigenicity [27,28]. The lipid antigens were purified by adsorption chromatography in an open Florisil column eluted with chloroform, acetone, and methanol (3 volumes each), followed by a solid phase extraction of the acetonic fraction onto a silica gel Sep-Pak column (Waters, USA) washed with chloroform and eluted with freshly distilled acetone. The purified lipids were dried under nitrogen and kept at −20 • C until use. To obtain specific hydrosoluble ligands against the purified lipid antigens, serum antibodies were raised in two white New-Zealand rabbits (2 months old). Briefly, the antigen-containing acetonic fraction was suspended in NaCl 0.85% (10 mg in 1.0 ml) and mixed with 1.0 ml of complete Freund adjuvant (Sigma Aldrich, USA). Each individual received a subcutaneous administration of 1.0 ml antigen cocktail. At 15 and 30 days post-immunization, each animal received boosts consisting of 3 mg antigen mixture in 0.5 ml 0.85% NaCl and 0.5 ml incomplete Freund adjuvant (Sigma Aldrich, USA). The subjects were bled before immunization to obtain Pre-immune, control serum (PS) (serum devoid of anti-lipid antibodies). A terminal bleed was performed at day 45 post-immunization to obtain a Hyper-immune serum (HS) (serum with high content of anti-lipid antibodies).

Sensor fabrication and sensing principle
In order to fabricate a PDMS end-capped FFPI sensor [29], we used a simple dip-coating technique. A mixture of PDMS precursor and cross-linker (10:1 by mass) was prepared, mixed, degassed and pre-cured at 70 • C during 15 min. This pre-curing process is essential to increase the viscosity of the PDMS drop in order to avoid Marangoni effects, which usually displace the polymer upwards along the fiber [30]. The polymer end-cap was realized by dipping the cleaved end face of a standard single-mode fiber (SMF-28e, Corning, USA) into a pre-cured drop of PDMS. A computer-controlled system for fiber positioning and pulling was used to set the appropriate velocities for driving the cleaved end-face of the fiber in and out of the pre-cured PDMS. Based on previous experiments for liquids analysis on optical fiber tips [30], the fiber displacement velocity was set to 1.7mm/s yielding a semi-spherical cap on the end face of the optical fiber. Afterwards, the end-capped sensor was vertically placed in an oven at 180 • C during 1 hour to fully solidify the PDMS. An image of the resulting fiber device is shown in Fig. 1(a). As shown schematically in Fig. 1(b),the PDMS end-cap and the cleaved end-face of the fiber form a FFPI. The spectral response of the FFPI can be described using two beam interference theory considering the Fresnel reflections from the two interfaces: the end-face of the fiber (core refractive index n 0 ) and the surface of the PDMS end-cap (reflection coefficients r 1 and r 2 , respectively). Both interfaces are separated by the PDMS cap, with length L 0 and refractive index n 1 . With this arrangement, the total reflected intensity from the FFPI (I R ) is given by [9,30]: where r 1 = n 1 −n 0 n 0 +n 1 , r 2 = n 2 −n 1 n 1 +n 2 , and the phase difference is: where λ is the wavelength of the incident light.
According to Eqs. (1) and 2, the intensity pattern I R is affected by variations of the refractive indices of the PDMS (n 1 ) and the surrounding medium (n 2 ), as well as by changes in the length of the resonant cavity (L 0 ). Therefore, when the bio-molecules are attached to the sensor surface, the interference pattern will shift in wavelength (∆λ) owing to changes in the effective length (∆L) and in the surrounding refractive index (∆n 2 ) [9,30]. The spectral shift can thus be associated with the modification of the PDMS surface due to the antigen attachment and further interactions of the sensor tip with the antibodies during the immunosensing test.

Characterization of the lipid-ligand system
For our experiments, a mixture of well-known lipid antigens was purified from M. fortuitum and used to produce anti-lipid antibodies in rabbits. The obtained antigen-antibody pair was then used to explore whether PDMS was suitable to immobilize a bioactive lipid (i.e. the mycobacterial cocktail antigen) and for promoting the capture of antibodies. Common protocols for lipid immobilization rely on the attachment of a very little amount of the lipid substance on the material surface; hence, a highly sensitive method is frequently necessary to assess the binding. Given the antigen nature of the lipids used in this report, an immune assay (i.e. indirect ELISA protocol [31]) with Pre-immune serum (PS) and Hyper-immune serum (HS) was performed on PDMS to evaluate the antigen-antibodies interaction.
Briefly, for the indirect ELISA protocol, the PDMS surface was coated by drop-casting with the lipid antigen diluted in ethanol-hexane (1:1, v/v); then the reactivity of the lipids to specific antibodies was measured with a secondary antibody linked to alkaline phosphatase, an enzyme for the development of color as a final signal. In this kind of assay, the values of the optical density (OD) are directly associated with the degree of antigen-antibody binding [31]. Different doses of rabbit sera were used to explore the ability of the functionalized PDMS to capture anti-lipid antibodies. Each serum was prepared in 0.3% BSA in PBS, using a log 2 serial dilution (1:100, 1:200, 1:400, 1:800, 1:1600, 1:3200, 1:6400, 1:12800), for a fixed unbiased antigen concentration of 64(µg/100µL). Similarly, to optimize the antigen concentration yielding the highest immunoreaction, the PDMS was functionalized by deposition of log 2 serial concentrations of lipid antigen onto the PDMS-containing wells, keeping fixed the optimum serum dilution obtained from the previous set of experiments. Hence, 300µL of the antigen were deposited at concentrations of 64, 32, 16, 8, 4, 2, 1, 0 (µg/100µL). All experiments were performed in triplicates.
The ELISA results for both set of experiments are shown in Fig. 2. As seen in the figure, mycobacterial lipid antigens readily allowed the capture of rabbit specific antibodies. In all cases, the reactivity of Hyper-immune serum (HS), rich in specific anti-lipid antibodies, was found to be larger than that of the control serum (PS). These data show that the lipid antigen maintain their bioactivity after attachment on the PDMS, thereby demonstrating the suitability of this polymer for immobilization of these molecules. From these plots we further obtain that a maximum capture of serum antibodies is achieved for a serum dilution of 1 : 400 ( Fig. 2(a)) and an antigen concentration of 8(µg/100µL) (Fig. 2(b)). Figure 3 shows the steps followed to functionalize and test the performance of the FFPI immunosensor. For PDMS functionalization, mycobacterial lipid antigens were immobilized on the end-capped sensor. This was achieved upon immersing the device in 0.5mL of ethanol-hexane (1:1, v/v) with an antigen concentration of 8(µg/100µL) for 30 minutes, followed by a 15 minutes drying process at room temperature (step 1). Free binding sites were blocked by immersing the sensor tip in 200µL of 3% BSA for 30 min at 37 • C (step 2), followed by three consecutive immersion washes in PBS (300µL) (step 3). A total of three sensors were functionalized and treated following this procedure.  Antibody binding on the functionalized PDMS end-cap was explored by immersing the devices in Pre-immune (PS) and Hyper-immune (HS) sera. Each sensor was first immersed in 200µL of pre-immune serum dilution (1 : 400(µL/µL)cuando) during 30 min at 37 • C (step 4). Subsequently, the same sensors were used for testing hyper-immune serum after removing the bound lipid ligands attached to the PDMS cap. This was done by immersion (3 times) in a denaturalizing buffer solution with pH 3 (step 5). The free binding sites on the fiber tips were immersed again in BSA for blocking (step 6) and finally washed three consecutive times in PBS (300µL) (step 7). Following this process, the tips were immersed in 200µL of 1 : 400(µL/µL) hyper-immune serum dilution during 30 min at 37 • C (step 8).

Functionalization of the FFPI sensor and direct immunoassay
While immersed in the PS and HS sera, the back-reflected interferometric signal at 1550nm from the sensors was continuously monitored using a fiber Bragg grating interrogator (Micron Optics, sm125) and a virtual instrument for automatic acquisition of the spectra. The spectral response (i.e. interference pattern) of the immunosensor was acquired every 10s during 30min. After these measurements were performed, the shift in wavelength (∆λ) over time was obtained by cross-correlation of the acquired spectra. This was done using the built-in Matlab algorithm and cross-correlating the reference signal (interference pattern acquired at t = 0) with the patterns acquired over time. The cross-correlation yields a maximum value indicating the corresponding wavelength shift, hence, the temporal evolution of ∆λ can be readily obtained.
Because the sera are kept under incubation conditions (37 • C), thermal effects are expected to contribute to the shift in wavelength registered by the sensors. This is mainly due to the thermal expansion coefficient of the PDMS (3.2 x 10 −4• C −1 ), that will lead to a change in the effective length (∆L) of the FP cavity. To account for this contribution, separate experiments were performed for registering the change in wavelength (∆λ) while the immunosensor was immersed in PBS at 37 • C during 30 minutes. Similarly, in order to verify that the trends in ∆λ obtained when testing PS and HS sera are only due to antigen-antibody interactions, untreated sensors (i.e., without the lipid antigen) were immersed in both sera. Under this conditions, no immuno-reaction should be expected and ∆λ should be the same for both sera.

Sensors features
As part of the characterization of the fabricated FFPI sensors, the spectral response of the device was monitored continuously during the different stages of the fabrication process. Spectra was thus acquired before PDMS deposition and until the curing of the PDMS end-cap was finished; images of the probe were acquired during the dip coating and after the curing process with an InfinyTube system coupled with a 10x microscope objective, keeping the same focal plane. Figure 4 shows the spectral response and images for these different stages. First, during the immersion of the tip on PDMS (case (a)), there is no interferometric signal since a cavity is not defined at this stage. Once the fiber is pulled out from the PDMS, the end-cap is formed and an interference pattern is clearly visible (case (b)). When the curing process starts, the Extinction Ratio of the interferometric signal (ER = P max − P min ) increases as a result of the higher refractive index within the cavity due to the thermal curing process [29]. While kept at a constant temperature (180 • C) for curing, the interference pattern experiences a blue-shift associated to a decrease in cavity length. This is noticeable in the spectra shown in (c) and (e) and is also apparent upon comparing the corresponding images. Finally, a comparison between the interference patterns displayed in (d) and (e), shows that no significant changes are apparent in the spectra. Hence, after a total curing time of approximately 25 min the PDMS end-cap is completely cured and the physical dimensions and refractive index remain constant.
Since all the relevant parameters for the dipping process are kept under control (i.e., initial volume of PDMS and pre-curing conditions, fiber pulling velocities and final curing of the end-cap), the fabrication procedure consistently yields sensors with similar characteristics. For all of the three sensors presented in this paper, the spectral features and their respective free spectral ranges FSR are similar (24.88 ± 0.02nm), corresponding to an effective cavity length (cap size) of approximately 33.42 ± 0.03µm.

Antigen immobilization and immunoassay results
Typical spectra obtained from our experiments showing the time evolution of the interference pattern for different stages of the immunoassay are shown in Fig. 5(a). For all the stages shown in the figure, the interference patterns shift over time towards longer wavelengths (red shift) while the tip is immersed in the solutions. These shifts correspond to an increase in the optical thickness due to the attachment of the bio-layer onto the PDMS end-cap. This is evident even from the functionalization process: the spectral features of the interference pattern of the sensor experiences both, a wavelength shift (∆λ = 4.28nm), a change in extinction ratio (∆ER = 0.53dBm) and a change in free spectral range (∆FSR = 0.1nm), thereby confirming the antigen immobilization on the PDMS. The shifts in wavelength over time (∆λ) obtained from cross-correlation of the acquired spectra for the different tested scenarios are plotted in Fig. 5(b). Notice first that thermal expansion effects are clearly visible from the sensor readout: after the sensor tip is immersed in PBS at 37 • C, ∆λ increases reaching a maximum after 30 minutes (curve A). At this time, the PDMS cap seems to reach its maximum length due to thermal expansion. For the FFPI without lipid antigens (curves B and C), a wavelength shift (∆λ i ) is initially registered upon immersion in the sera (either PS or HS) owing to the difference in refractive indices of these solutions and the PBS [32][33][34]. Once immersed in the PS and the HS solutions, ∆λ has a very similar trend to that obtained in PBS (curve A). This trend in wavelength shift is thus attributed to thermal expansion effects leading to an increase in length of the PDMS cavities. We can thus conclude that there is no attachment of antibodies on the FFPI without the antigen; thus, neither the PS nor the HS can be identified properly. Furthermore, this confirms that immunosensing is not achieved through polymer-antibodies interactions, and that as shown below, the antigen functionalization is therefore essential for antibody detection.
The wavelength shifts obtained with the functionalized sensors when immersed in both sera (PS and HS) are shown in curves D and E of Fig. 5(b). While ∆λ still changes over time showing the thermal expansion effects of PDMS, the maximum wavelength shifts obtained with both sera (PS and HS) are larger than those registered for the other devices (curves D and E). Furthermore, the trend of these curves is different to those obtained for curves A, B and C, which are driven mostly by thermal expansion effects. This suggests that antibody binding on the functionalized PDMS cap indeed has an effect on the spectral response of the sensors. Notice further that the ∆λ registered during immersion in the HS serum (curve E) is larger than that obtained with the PS serum (curve D) (∆λ HS = 4.95nm vs. ∆λ PS = 4.03nm). This is a reasonable outcome as the HS serum contains a larger number of specific antibodies, thus increasing the binding events and thereby leading to a larger ∆λ. It is also important to remark that the readouts for the HS serum were collected after cleaning and blocking the probes once the results for PS were obtained. This therefore confirms that the antigen coating is preserved on the PDMS cap and the sensors can be readily reused after simple protein denaturation washes. Thus, the capabilities of the immunosensor based on PDMS functionalized with lipid ligands seem to scale well to suitable specific antibody detection and serum differentiation with attractive features (e.g. label-free and reusable system with real-time and direct detection, and facile fabrication and functionalization process). Table 1 shows a summary of our results. For comparative purposes, we can estimate the difference between the maximum wavelength shift (∆λ f ) and the initial shift registered during immersion of the devices (∆λ i ). This difference (S ∆λ = ∆λ f − ∆λ i ) represents the effective shift in wavelength due to thermal effects and any antigen-antibody interactions. Clearly, the largest S ∆λ is achieved for the Hyper-immune (HS) serum and further shows a marked difference with the Pre-immune (PS) (0.92nm). Thus, the FFPI immunosensor based on PDMS functionalized with bioactive lipids is able to properly differentiate between different antibody-containing sera. The wavelength shift occurring during the immunoassay test can be estimated as [9]: Considering the constributions by the bio-layer and the thermal effects in the PDMS, the changes in length and refractive index (∆L and ∆n) can be expressed as: where ∆T is the increase in temperature, α L = 3.2x10 −4 1 • C and α n = −1.8x10 −4 1 • C correspond to the thermal expansion and thermo optic coefficients of PDMS, respectively [35,36]. Using these values and our experimental conditions (L 0 = 33.42µm, n 1 = 1.3997 [37], ∆T = 37 • C − 23 • C = 14 • C, and λ = 1550nm), Eq. (3) can be rewritten as: ∆λ = 3.04nm + (46.38x10 −3 )L bio + (1.1x10 −6 )n bio (6) The first term in this expression represents the thermal effects contributing to the shift in wavelength. Notice that this value agrees well with the S ∆λ obtained for the untreated sensors (i.e., without the antigen, curves A, B and C). In these cases, there is no antigen-antibody interactions and only thermal effects are expected to produce a shift in wavelength. For the functionalized sensors (curves D and E), the wavelength shifts due to bio-recognition effects might be evaluated upon substracting the thermal effects contributions (i.e., S ∆λ − 3.53nm). This yields 0.50nm and 1.42nm for the Pre-immune (PS) and the Hyper-immune (HS) serum, respectively. According to Eq. (6) the contribution of the bio-layer thickness is much larger than its refractive index. Thus, neglecting the last term in this equation we can estimate that the bio-layer attached to the immunosensor for pre-immune serum has a thickness of 10.78nm, while for hyper-immune serum reaches 30.61nm. Note that these layers include the antigen coating and any antibodies that could have attached to the functionalized PDMS surface. Hence, considering that the tests were performed with the same immunosensors, the thickness difference could be attributed to changes in the PDMS end-cap surface owing to antigen-antibody binding effects, as resported previously for other bio-sensing platforms [9,38].

Conclusions
We have demonstrated the detection of antibodies with a fiber optic sensor based on PDMS and bioactive lipids. Our results show than the proposed FFPI immunosensor is able to differentiate antibody-containing sera through changes in the spectral features of the device. These are associated to the interactions between the active lipids and the serum antibodies. The FFPI immunosensor shows attractive features such as ease of fabrication, label-free operation, real-time monitoring capabilities and reusability. Because sensing is based on antigen-antibody interaction, the end-capped devices may be used for detection of any type of target molecule, provided a proper antigen is immobilized on the PDMS. To the best of our knowledge, this is the first report showing that PDMS holds good functionalization capabilities by simple deposition of lipid mediators. Evidently, further studies are needed to assess the usefulness of PDMS for the immobilization of other bioactive lipids. Nonetheless, our results demonstrate a novel means for fabrication of optical biosensors for lipidomics tools analysis, such as those needed for the screening of unknown ligands or their reactivity with human fluid components (biomarkers of diseases, for instance) which require the availability of supports for bioactive lipids attachment. Indeed, the exploration of PDMS functionalized with mycobacterial lipid antigens, such as the one presented here, is highly encouraging for the diagnosis tuberculosis, currently one of the top killer diseases worldwide.