Comparison of antibody immobilization on renewable nanocrystals for paper-based detection

1 Introduction

Cancer remains a leading cause of morbidity and mortality worldwide, and its burden continues to rise as populations grow and age. Global estimates for 2022 indicate ∼20 million new cases and ∼9.7 million deaths, with a lifetime risk on the order of 1 in 5 of developing cancer (Belfakir et al., 2024; Bray et al., 2024). These trends reinforce a persistent clinical need: earlier, more accessible detection that can be deployed beyond centralized laboratories and that can support screening, triage, and longitudinal monitoring in resource-variable settings.

Biomarker-centric diagnostics are central to this agenda, yet performance at the point of need is often constrained by the capture interface. The surface chemistry and microstructure that governs probe loading, orientation, and non-specific adsorption are often the most relevant but understudied factors. For epithelial malignancies, carcinoembryonic antigen-related cell adhesion molecule 5 (CEACAM5, also commonly referred to as CEA and CD66e) is a widely used target, and it has served as a workhorse analyte for developing and benchmarking biosensor architectures (Blumenthal et al., 2007; Hatakeyama et al., 2013). Despite extensive work on CEA detection, translation of “high-sensitivity” benchtop designs into robust low-cost formats remains limited by matrix effects, inconsistent antibody immobilization, and variable signal-to-noise when assays are implemented on porous substrates or minimally instrumented platforms (Xu et al., 2020).

Paper-based analytical devices and paper microfluidics are attractive candidates to close this gap because they can satisfy many of the operational criteria expected of point-of-need testing (low cost, portability, minimal user steps, and compatibility with distributed manufacturing) (Pradela-Filho et al., 2023). However, the same attributes that make paper attractive (porosity, capillarity, and heterogeneous fiber chemistry) create non-trivial surface-science problems for immunoassays: (i) uncontrolled physical adsorption can denature proteins or bury binding sites; (ii) covalent coupling can be inefficient without precise control of accessible functional groups; and (iii) blocking and antifouling strategies must be balanced against mass-transfer limitations intrinsic to porous networks (Belfakir et al., 2024; Herrera et al., 2025; Holstein et al., 2015). Consequently, the immobilization strategy is frequently the dominant determinant of analytical performance (capture capacity, apparent affinity, kinetics, and background), particularly when assays must operate in complex fluids.

Recent analyses of cellulose-based immunoassay surfaces emphasize that no single immobilization route is universally optimal; instead, outcomes depend on the interplay between cellulose oxidation state (e.g., aldehyde/carboxy group density), coupling chemistry (e.g., 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide/EDC with N-hydroxysuccinimide/NHS, or aldehyde coupling via glutaraldehyde/pentane-1,5-dial), protein microenvironment, and the porous transport regime (Belfakir et al., 2024; Holstein et al., 2015). In parallel, there is growing interest in using bio-based nanocrystals as “interfacial amplifiers” to increase surface area, tune charge/hydrophilicity, and provide higher-density, better-controlled antibody presentation than untreated cellulose fibers alone. Cellulose nanocrystals (CNCs) have emerged as versatile platforms for bio interfaces and sensing, with strong emphasis on functionalization routes that preserve biorecognition while enabling scalable processing (Liu et al., 2024; Nori et al., 2023).

An underexplored opportunity is to compare CNCs against structurally analogous, renewable nanocrystals with distinct surface chemistries, such as chitosan nanocrystals (ChNCs) and CNC/chitosan (CNC/Ch) composites (Gomez-Maldonado et al., 2024). Chitosan-derived nanocrystals introduce primary amino-rich surfaces that can be protonated under mildly acidic conditions, enabling direct electrostatic coupling to negatively charge biomolecules and providing tunable cationic character for controlled interfacial assembly. This provides orthogonal control over electrostatics, hydration, and potential protein orientation relative to sulfated CNCs. In buffered saline media, however, the apparent surface charge of ChNCs can be substantially screened by ionic strength and specific ion pairing, so electrostatic contributions may manifest primarily through hydration/structural effects rather than high ζ-potential values. ChNC-stabilized biopolymer architectures have also shown functional benefits in aqueous systems relevant to biosensing (e.g., enhanced stability and tailored interfacial properties), motivating a systematic evaluation of ChNC-containing platforms for immunoassay capture layers (Manna et al., 2026).

Importantly, a comparative framework is still missing: most studies report a single nanomaterial/chemistry combination under a single assay condition, which makes it difficult to disentangle whether gains arise from increased surface area, altered charge, improved antibody conformation, or reduced non-specific adsorption. The aim of this work is to integrate physicochemical characterization with real-time interfacial quantification (QCM-D and MP-SPR) to separate material effects (surface chemistry, charge, hydration) from functional effects in antibody-antigen interactions (affinity proxies and non-specific adsorption). The working hypothesis of this work is that the different nanocrystals will exhibit variation in surface chemistry (carboxyl vs. amino functionality), charge density, and hydration behavior, and therefore, different antibody immobilization and interfacial organization are expected. This approach enables the establishment of actionable design rules for paper-based immunoassays targeting CD66e/CEACAM5 and related biomarkers.

2 Experimental

2.1 Materials

Sulfuric acid was purchased from Fisher Scientific (Hampton, NH, USA). Chitosan from shrimp shells with a degree of acetylation (DA) of 19% was received from Vanson (Lot no: VNS-389, Redmond, WA, USA). The ionic liquid 1-butyl-3-methylimidazolium hydrogen sulfate ([C4mim][HSO4]) was obtained from ProIonic GmbH (Graz, Austria). Milli-Q deionized (DI) water (conductivity ∼4.2 MΩ·cm) used for precipitation and purification was produced using a Millipore Milli-Q system (Burlington, MA, USA). N-hydrosuccinimide (NHS) was obtained from Spectrum Chemical, MFG Corporation (Gardena, CA, USA), and 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) was purchased from Chem-Impex International (Wood Dale, IL, USA). For the composite, 85% deacetylated chitosan (Catalog #J64143.30 Lot. R15K037) from Thermo Fisher Scientific was used (Waltham, MA, USA). Phosphate Buffered Saline Tablets (PBS) were sourced from Research Products International (Mount Prospect, IL, USA). The biological reagents, Monoclonal Antibody CD66e (Catalog # 1820-4700) and Carcinoembryonic antigen (CD66e) (Catalog # PHP282), were purchased from Bio-Rad Laboratories, Inc. (Hercules, CA, USA) at concentrations of 1.0 mg/mL. Whatman Grade 41 filter paper was purchased from Cytiva (Little Chalfont, Buckinghamshire, UK). Gold nanoparticles (GNP, citric acid stabilized, 40 nm) were acquired from nanoComposix (San Diego, CA, USA). Polyvinyl sulfuric acid potassium salt (0.001 N, PVSK) and polydimethyl diallyl ammonium chloride (0.001 N, pDADMAC) were bought from BTG (Eclépens, Switzerland). Substrates consisted of P-type Test Grade silicon wafers (thickness 500 µm) purchased from University Wafer MS (South Boston, MA, USA).

2.2 Cellulose nanocrystals isolation

The CNCs were prepared according to a previous study (Hu and Abidi, 2016). Raw cotton linter was harvested from cotton plants locally (Lubbock, TX, USA). Prior to its use, scouring and bleaching procedures were performed to remove impurities to obtain highly purified cotton fiber containing 99% pure cellulose.

Purified cotton linter (11 g) was hydrolyzed in 100 mL of 63.5% (v/v) H₂SO₄ at 45 °C for 90 min. The resulting suspension was centrifuged at 3,074 × g for 15 min to separate the supernatant from the pellet. The supernatant was discarded and the pellet was washed with DI water. The centrifugation was repeated at the same speed until the supernatant was no longer clear. The milky supernatant was then carefully decanted and neutralized with aqueous KOH (3 mol/L), followed by dialysis for 3 days. After dialysis, the suspension was transferred to a glass vessel and was allowed to stand for 3 days. The top layer was collected and ultrasonicated using a Vibra-Cell ultrasonicator (Model CV 33, Newtown, CT, USA) for 5 min prior to further use.

2.3 Chitosan nanocrystals preparation

For ChNC, 5 g of dried chitosan was mixed with 50 g of [C4mim][HSO₄] ionic liquid (IL, 9 wt% loading) in a 250 mL round-bottom flask. The mixture was thoroughly homogenized using a glass rod and subsequently placed in an oven maintained at 45 °C–65 °C for 24 h to ensure complete blending and effective pretreatment of chitosan in the ionic liquid. After the pretreatment stage, the mixture was removed from the oven, and DI water (∼30 wt% relative to the ionic liquid) was added. The resulting suspension was then hydrolyzed at 100 °C for 24 h on an IKA hot plate (Wilmington, NC, USA) equipped with a condenser under continuous magnetic stirring. Following hydrolysis, the reaction mixture was allowed to cool to room temperature, after which 1 M NaOH was added dropwise to neutralize the acidic ionic liquid, inducing precipitation of the hydrolyzed polymer from the solution. The suspension was transferred into 50 mL centrifuge tubes and washed repeatedly by alternating stirring and centrifugation (Sorvall X Pro/ST Plus Series centrifuge, Osterode am Harz, Germany) for 12–15 cycles, at 15,316 × g with DI water to remove residual ionic liquid until the supernatant reached neutral pH. The recovered material was subsequently ultrasonicated using a Vibra-Cell ultrasonicator (Model CV 33, Newtown, CT, USA) for approximately 15 min to obtain a uniform suspension, which was used directly in subsequent processing steps.

2.4 CNC-chitosan composite formation

The composite was generated as in our previous work (Gomez-Maldonado et al., 2024). A 2% (w/v) chitosan (Ch) solution was prepared by dissolving 0.4 g of chitosan in 20 mL of 1% (v/v) acetic acid under continuous stirring overnight at room temperature. In parallel, a 0.2% (w/v) cellulose nanocrystal (CNC) suspension was prepared in DI water. The CNC suspension was transferred to 35 mL centrifuge tubes and centrifuged at 11,200 × g for 45 min. The supernatant was then carefully removed. The resulting pellet was resuspended in the prepared chitosan solution and vortexed to promote interaction between chitosan and CNCs.

The CNC/Ch suspension was transferred to a 50 mL beaker and subjected to tip sonication using Vibra-Cell ultrasonicator (Model CV 33, Newtown, CT, USA) for 10 min using 30 s on/off cycles at 20% amplitude, with a total energy input of 30 kJ. The suspension was then stirred overnight at room temperature. To further purify the CNC/Ch complex, the suspension was centrifuged at 11,200 × g for 30 min. The pellet was resuspended in DI water, vortexed, and centrifuged again under identical conditions. Finally, the pellet was resuspended in 3 mL of 1× phosphate-buffered saline (PBS) and sonicated for 10 min (30 s on/off cycles, 20% amplitude, 30 kJ). The final CNC/Ch suspension was stored at 4 °C until further use.

2.5 Antibody immobilization

CNC/Ch (0.9% w/v in 1X PBS), CNC (0.9% w/v in 1X PBS), and ChNC (0.9% w/v in 1X PBS) suspensions were transferred to separate 50 mL reaction flasks. To each sample, 0.2 µL (0.2 µg) of mouse anti-human CD66e antibody (initial concentration of 1.0 mg/mL) was added, followed by 5 µL of 50 mM 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) and 5 µL of 200 mM N-hydrosuccinimide (NHS). The reaction mixtures were allowed to proceed under continuous stirring at room temperature for approximately 12 h.

The reactions were quenched by the addition of 22.4 µL of ethanolamine. Samples were then centrifuged at 9,000 × g for 1 h, after which the supernatants were discarded. The resulting pellets were resuspended in 3 mL of 1X PBS, vortexed thoroughly, and centrifuged again under identical conditions to perform a second wash. Antigen control dot for paper-based testing was performed following the same procedure but increasing the antigen concentration to 2 µL (0.2 µg) of CD66e antigen.

2.6 Characterization techniques

2.6.1 Atomic force microscopy (AFM)

Atomic Force Microscopy was used to analyze the morphology of the different nanocrystals. For this purpose, an Asylum Research MFP-3D-SA Atomic Force Microscope (Goleta, CA, USA) was used, along with an Asylum Research cantilever model AC160TS (Goleta, CA, USA). Images were obtained using Igor Pro 6.37 in AC Air Topography mode. 2µm × 2 µm scanning areas were scanned to capture height, amplitude, and phase. Furthermore, the image correction was enabled using Gwyddion 2.61.

2.6.2 Dispersion behavior by dynamic light scattering (DLS)

The hydrodynamic radius and ζ-potential of the samples were obtained for 0.01 wt% aqueous dispersions prepared using PBS as the dispersion medium, which were previously adjusted to pH 6.5 by the dropwise addition of dilute NaOH or HCl solutions. Measurements were performed using an Anton Paar Litesizer DLS 500 (Graz, Austria) at 25 °C. Prior to analysis, the samples were vortexed and sonicated in a water bath.

2.6.3 Fourier transform infrared with attenuated total reflection (FTIR-ATR)

The chemical structure characterization of the base and modified materials was performed using a Bruker ALPHA II FTIR system equipped with a diamond crystal (Billerica, MA, USA). To eliminate water readings, the samples were freeze-dried overnight prior to measurement using 10–20 mg. Spectra were recorded in the spectral range of 4000–400 cm^-1 with 2 cm^-1 resolution (maximum given for the equipment) and 128 scans for each sample. FTIR data were normalized, and the baseline was corrected using the software OPUS 9.0.6.

2.6.4 Polyelectrolyte titration

Charge density was determined by colloidal titration using the CHEMTRAC Laboratory Charge Analyzer LCA5 (Norcross, GA, USA). The aqueous dispersion of the samples was adjusted to 0.04 wt% with a pH of 6.5. Depending on the original charge of the media, 25 mL of p-DADMAC (0.001N) or PVSK (0.001N) was added to 15 mL of the sample dispersion. The mixture was well mixed and centrifuged to remove any aggregates. 10 mL of the supernatant was collected and titrated against the opposing polymer (0.001N PVSK or p-DADMAC) in the charge analyzer, per triplicate. The charge density was calculated based on the volume required to reach a streaming current value (SVC) of 0 using WebPlotDigitizer, to extract numerical data from images obtained from the CHEMTRAC.

Charge density = p D A D M A C * V p D A D M A C − P V S K * V P V S K W d r y s a m p l e

Where:

[pDADMAC] is the concentration of the cationic polymer.

V_pDADMAC is the used volume of pDADMAC.

[PVSK] is the concentration of the anionic titrant.

V_PVSK is the consumed volume of titrant.

W_{dry sample} is the weight of the dry sample.

2.6.5 Potentiometric titrations

Titration to quantify the main available functional groups was performed by measuring pH and conductivity with a Mettler Toledo Seven Direct SD23 pH and conductivity meter (Columbus, Ohio, USA). The conductivity values were corrected for the dilution effect carried by the progressive addition of NaOH during the titration. Therefore, a volume correction was applied to normalize the conductivity values to the initial suspension volume, enabling an accurate tracking of the changes associated exclusively with the neutralization of the functional groups on the material surface. This correction was done according to the following equation:

C o n d u c t i v i t y c = C o n d u c t i v i t y m × V i + V o V o

Where:

Conductivity_c is the corrected conductivity (μS cm^-1).

Conductivity_m is the measured conductivity for each data point (μS cm^-1).

V_i is the initial suspension volume (mL).

V_o is the added volume of NaOH at each point (mL).

2.6.5.1 Sulfate half-ester titration

Sulfate half-ester groups introduced onto the surface of the CNCs during the hydrolysis reaction were quantified by conductometric titration. CNC samples were dialyzed (Spectra/Por 4 membrane, 12–14 kDa MWCO) for at least 5 days until the conductivity did not change to eliminate any free residual sulfate ions. 120 mg of dialyzed sample was diluted to a final volume of 100 mL with ultrapure water, followed by protonation through the addition of 1 mL of HCl (0.05 M) to reach a pH around 3.5 and 300 μL of NaCl (0.1 M) to increase the conductivity values. The sample was titrated with NaOH (0.05 M) added in 100 μL increments. The resulting curves showed an initial decrease in conductivity associated with the neutralization of protons bound to sulfate ester groups, followed by an increase once all charged groups were neutralized and excess hydroxide ions were present. The equivalence point, identified from the inflection of the curve, was used to determine the volume of NaOH at a known concentration required for complete neutralization.

C R − O S O 3 H m m o l g = C N a O H × V 2 − V 1 m s a m p l e × 1000

Where:

C _R-OSO3H is the sulfate half-ester group content (mmol/g).

C_NaOH is the concentration of the NaOH titrant (mol/L).

V₁ is the volume of NaOH required to neutralize excess HCl (L).

V₂ is the total volume required to neutralize both HCl and sulfate half-ester groups (L).

m_sample is the dry mass of the sample used for titration (g).

2.6.5.2 Carboxy titration

The quantification of carboxy groups was executed for the CNC and CNC/Ch samples, where 20 mg of the sample was dissolved in 10 mL of HCl (0.01 M). After 10 min of stirring, the suspensions were titrated with aqueous NaOH (0.01 M) in portions of 500 μL each time and 100 μL closer to pH 9-11. The conductivity was measured and corrected after every addition of NaOH solution. The titration proceeded over the pH range of 2–12. The conductimetric curve showed the presence of a strong acid, corresponding to excess HCl, and a weak acid, corresponding to the content of carboxy groups.

C C O O H m m o l g = C N a O H × V 2 − V 1 m s a m p l e × 1000

Where:

C_COOH is the carboxy group content (mmol/g).

C_NaOH is the concentration of the NaOH titrant (mol/L).

V₂ −V₁ is the volume of NaOH required to neutralize the carboxy groups, determined from the conductivity minimum (inflection point) (L).

m_sample is the dry mass of the sample used for titration (g).

2.6.5.3 Amino titration

The quantification of amines was performed on the CNC/Ch and ChNC samples, where 50 mg of the sample was added to 10 mL of HCl (0.05 M) and left to stir for 10 min. The mixture was then titrated with NaOH (0.05 M), adding 0.5–1 mL at a time until the equivalent point was reached. Based on the titration curve, the inflection point was determined, and degree of functionalization calculated using the following equation (Moaseri et al., 2013).

D e g r e e o f f u n c t i o n a l i z a t i o n = V 1 m L * C 1 M − V 2 m L * C 2 M S a m p l e w e i g h t g

2.7 Adsorption experiments

2.7.1 Quartz crystal microbalance with dissipation monitoring (QCM-D)

Titanium/gold (Ti/Au) coated quartz crystal sensors (AWSensors, Valencia, Spain) were used in all experiments. Prior to use, sensors were cleaned for 30 min using a Novascan PSD Series digital UV–ozone system (Milwaukee, WI, USA). Sensor surfaces were subsequently activated by immersion in a 0.1 wt% polyethyleneimine solution for 20 min, followed by rinsing in ultrapure (UP) water for 5 min. Sensors were coated with 180 µL of a 0.01% (w/v) solution of each sample using a Laurell H6-23 spin coater (Lansdale, PA, USA) at 4500 rpm.

Surface functionalization and antigen capture were monitored using a BioNavis QCMD 400 system (Tampere, Finland). QCM-D was used to evaluate the adsorption behavior and viscoelastic properties of antibody-conjugated Ch-CNC, ChNC, and CNC coatings deposited onto Ti/Au-coated sensors. Prior to QCM-D measurements, the coated sensors were rehydrated in 1X PBS for 24 h to allow the films to reach equilibrium under hydrated conditions. This step was critical to minimize non-specific responses and to avoid sudden swelling or structural rearrangements upon buffer introduction, which could otherwise lead to false positive or negative signals.

Following surface functionalization, the response of each coating to antigen binding was investigated by exposing the sensors to different antigen concentrations. Based on previous work, effective antigen binding was sensed using a solution prepared by diluting 4 µL of a stock antigen solution into 5 mL of PBS, corresponding to a concentration of 0.8 ng/μL (Liu et al., 2024; Nori et al., 2023). This concentration was selected as the reference (100%) to prepare the subsequent antigen solution, corresponding to 10%, 50%, 200%, and 400% of the initial antigen solution.

Changes in frequency (ΔF/n) and dissipation (ΔD) were monitored in real time to assess mass uptake and mechanical properties of the resulting films. Antigen solutions at varying dilutions were introduced into the flow cells for 30 min. Each sample was analyzed in quadruplicate. All experiments were performed at 25 °C under a constant flow rate of 50 μL/min. Although data were collected for all overtones, only the third overtone is presented for clarity.

Frequency and dissipation changes were quantified by exporting the full sensogram data to Microsoft Excel (Version 16.104.1), with values recorded at 1-s intervals. For each measurement cycle, the ΔF/n and ΔD values were averaged over the stable PBS baseline regions before and after antigen exposure. The frequency and dissipation shifts associated with each antigen concentration were then calculated as the difference between consecutive PBS baselines (i.e., PBS_post–PBS_pre), rather than directly from the antigen exposure step. This approach was to ensure that the (irreversible) attachment was quantified, avoiding transient effects during antigen flow and permitted reliable comparisons of antigen-induced changes across concentrations and coating types. The adsorbed mass was calculated from the normalized frequency shifts using the Sauerbrey equation, assuming rigid, evenly distributed films.

C h a n g e i n m a s s = − C · Δ F n

Where:

C = 17.7 ng·cm^-2·Hz^-1 for a 5 MHz crystal.

ΔF = Change in resonance frequency

n = harmonic number (overtone).

2.7.2 Multiparametric surface plasmon resonance (MP-SPR)

The MP-SPR sensors were prepared following the same procedure described for the QCM sensors, including the stabilization in PBS. Antigen adhesion was quantified by injecting a series of dilutions (0%, 10%, 50%, 100%, 200%, and 400%) using an MP-SPR Navi™ 210A VASA system from BioNavis (Tampere, Finland). Each injection was performed for 10 min (total injected volume of 340 µL), followed by a 5 min post-injection delay with PBS to stabilize the system before the subsequent injection. Measurements were conducted at 25 °C under a constant flow rate of 20 μL/min.

Data analysis was performed using the 980 nm laser only. All samples were measured in quadruplicate. The reported data is the difference between the PBS baseline and the stable stage of the rinsing prior to the subsequent injection. Sensograms were exported and analyzed in Microsoft Excel (Version 16.104.1) to quantify the changes in the SPR signal associated with antigen binding, and the stabilized PBS baseline before and after each antigen exposure (ΔSPR WeightedCentroid Angle). Measurements were performed in quadruplicate and averaged for subsequent analysis.

2.8 Paper-based detection

2.8.1 Gold nanoparticle (GNP) detection probe formation

Conjugation was performed as described by nanoComposix for the Passive Conjugation Protocol BioReadyTM 40 nm Bare Gold with a modification duplicating the concentration proportionally. For this, 25 µL of PBS adjusted at pH 9 were added to an Eppendorf tube, mixed with 20 µL of antibody (0.2 mg/mL) and 500 µL of GNP. The resulting mixture was vortexed, covered with aluminum foil to protect it from light and incubated on the Shaker Solaris Orbital from ThermoScientific at 0.3 × g with a 25 mm orbital diameter for 30 min. After incubation, the conjugate was centrifuged at 19,000 × g for 1 h at 3 °C to obtain the pellet and resuspended with 250 µL of PBS.

2.8.2 Device assembly and testing

Filter paper was sectioned to produce an assay with three equal flaps, resulting in three replicates per sample in a similar structure to previous work (Gomez-Maldonado et al., 2024). For each test, 1 µL of 1% of Tween-20 in PBS was deposited on the edge of the test strip. Then, 1 cm from the surfactant, 1 µL of the test sample was deposited 3 times, allowing each drop to dry completely before the subsequent application. The immobilized antigen for the control line was prepared following the procedure described in Section 2.5. Subsequently, 1 µL of the solution was deposited three times at 1 cm from the test line, allowing the spot to dry between each deposition. The assay was allowed to dry for 5–10 min at ambient conditions.

The filter paper was then positioned in an immersion solution containing GNP conjugated (54 μL, OD₅₂₀ = 1.45 ± 0.01), 5 µL of antigen, and 200 µL of PBS. The assay proceeded for 2–5 min, during which the GNP migrated up the filter paper to interact with the samples and antigen dots. After the assay was done, the filter was left to dry, and the results were evaluated. Pictures were taken with a Nikon D3200 (18–55 mm) under homogeneous light.

2.8.3 Image analysis

Images were processed using Matlab 2025b (Mathworks, Inc., Portola Valley, CA). On each flap of the filter paper, bounding boxes of equal size were formed around the locations of the drops for both the control and the test solution. Also, a bounding box was formed in the region between the two drop locations to represent the background. This concept is illustrated in Figure 1. In that figure, the blue boxes indicate the control regions, the green boxes indicate the test regions, and the yellow boxes indicate the region sampled as a representative background. Each bounding box region was cropped and converted to the CIELAB color space, and the mean L, a, and b* values were computed to obtain a single representative CIELAB vector. The visual impact of the solution was quantified by evaluating the CIEDE2000 color difference between the vectors of the solution region (control or test) and the corresponding background region (Sharma et al., 2005). This effectively provides a quantitative color difference for each of the two solution regions versus the surrounding area. The differences were averaged across the three reps to yield a single average color difference for the control lines and a single average difference for the test lines.

2.9 Statistical analysis

Statistical analyses and graphing were performed using GraphPad Prism (version 10.6.1; GraphPad Software, San Diego, CA, USA). Chemical differences among all materials were analyzed through one-way analysis of variance (ANOVA). When a significant effect was observed (p < 0.05), Tukey’s multiple comparisons test was applied to identify statistically significant differences between individual materials. Pairwise comparisons within the same material before and after antibody immobilization (e.g., CNC vs. CNC-Ab) were performed using an unpaired t-test with Welch’s correction. Antigen binding analyses were performed by Friedman’s ANOVA to determine statistically significant differences between samples. For the paper-based devices results, statistical comparison was done using a Kruskal–Wallis test for nonparametric data, with the pairwise comparison done with a corrected Duncan test. For all statistical analyses, a p-value <0.05 was considered significant.

3 Results

3.1 Renewable nanocrystals comparison and immobilization confirmation

Morphological characterization of the nanocrystals was done with the aid of AFM, while their hydrodynamic diameter and colloidal stability in PBS were followed by DLS. As shown in the microscopy images in Figure 2, CNCs had a nanoscale size, with visible aggregation. The presence of aggregates was confirmed by DLS measurements (Table 1), which reflected aggregated structures in PBS rather than the primary nanocrystal dimensions. The measured hydrodynamic diameter was 1783 ± 437 nm and a ζ-potential of −22.1 ± 0.7 mV, which is outside of the values considered as colloidally stable. Once conjugated with chitosan, to form CNC/Ch, there was a visible increase in the dispersion of the material; however, the hydrodynamic value was not significantly different from the values obtained from the pristine CNCs. Nevertheless, there was a flip in the ζ-potential values to a positive 1.8 ± 1.5 mV, confirming the presence of the chitosan in the surface. A smaller particle size of 293.4 ± 6.5 nm was measured for the ChNC, which is also perceptible in the AFM images, where fewer aggregates are observed and a more globular like structure was measured. Despite this apparent better dispersion, their ζ-potential still had values outside of the colloidal stability range. Particularly for ChNC, in PBS at pH 6.5, ζ-potential values can remain low due to double-layer compression and counterion screening even when titratable amino groups are present.

TABLE 1. Physicochemical characterization of the nanocrystals before and after antibody immobilization.

Different lowercase letters (a, b, or c) within the same column indicate statistically significant grouping among the materials (p < 0.05). An (*) denotes statistically significant differences between the same material before and after antibody immobilization (p < 0.05).

In contrast, after antibody immobilization, there was a noticeable morphological change for all the tested changes observed after antibody conjugation. They no longer displayed an elongated, fibrillar structure, but instead appeared more agglomerated, likely due to protein adsorption on their surfaces. Significant changes in particle size for the CNCs and CNC/Ch after immobilization confirmed antibody immobilization on the surface. The CNCs almost duplicated their hydrodynamic diameter after antibody attachment, while decreasing their ζ-potential to almost two-thirds of their initial values. Conversely, the CNC/Ch composite seemed to have improved their dispersibility after modification, to almost half their initial values (from 4466 ± 153 to 2535 ± 192 nm), with their ζ potential decreasing its variance range. However, despite their improved dispersibility, for CNC/Ch-Ab, aggregation upon drying was perceptible in the AFM images, where larger aggregates and an irregular morphology were observed. On the other hand, the ChNC did not exhibit statistically significant changes in hydrodynamic radius or ζ-potential after antibody immobilization; however, the broadening of the hydrodynamic diameter distribution suggests heterogeneous antibody coverage on the particle surface. This increase in distribution was further confirmed using AFM images, which showed more aggregates of 2-3 nanoparticles than the unmodified ChCN. Interestingly, the ChNC-Ab seemed to have a more elongated shape after antibody immobilization, suggesting conformational changes after the processing. It is worth remembering that the difference between the values on DLS and AFM can be explained by the difference in the technique principles, DLS measures the hydrodynamic diameter of the swollen suspended particles, while AFM shows particles that are immobilized to the surface and dried prior observation.

The trends and changes observed were then corroborated by performing chemical characterization of the nanocrystals, both prior to and after antibody immobilization. This was done by FTIR (Figure 3), and different conductivity-based titrations, both ionic and polyelectrolyte (Table 1).

The qualitative changes measured with the FTIR spectra showed that for the pristine CNCs, the characteristic absorption bands of cellulose are present, including a broad O–H stretching band at 3300–3400 cm^-1, C-H stretching vibration ∼2890 cm^-1, C−O stretching, and C−H rocking vibrations in the CNC pyranose rings lead to absorption bands at ∼1000–1100 cm^-1 (Wei et al., 2024). The composite (CNC/Ch) preserved the characteristic absorption bands of both polysaccharides. A prominent band in the 2850–2950 cm^-1 region is observed, which is assigned to C–H stretching vibrations of aliphatic groups from the polysaccharide backbones of CNC and ChNC (Gomez-Maldonado et al., 2024). ChNC exhibits a broad band at 3372 and 3301 cm^-1 corresponding to overlapping O–H and N–H stretching vibrations, C–H stretching at ∼2876 cm^-1, a prominent band at ∼1630 cm^-1 is primarily attributed to the amide I (C=O) vibration of residual N-acetylated units in chitosan (consistent with the reported DA); however, overlapping with contributions from adsorbed water (H–O–H bending) have to be considered. N–H bending of–NH₂ groups are present at 1515 cm^-1. The amide III band (C–N and N–H) appears at 1375 cm^-1, C–O stretching vibrations at 1025 cm^-1, along with characteristic C–O–C vibrations resembling the glycosidic linkages at 1152 cm^-1 and 1068 cm^-1 (Panackal Shibu et al., 2025).

After antibody association (CNC-Ab), subtle changes in the O–H stretching region were detected, suggesting an increased hydrogen-bonding interaction involving surface hydroxy groups. Similarly, only minor variations in this region are observed after antibody association for the CNC/Ch-Ab composite. While the spectrum of the ChNC after antibody immobilization (ChNC-Ab), presented subtle changes in the intensity and shape of the O–H/N–H stretching region and the amide bands were observed, suggesting changes in hydrogen bonding and amide/amino band intensities consistent with protein association at the ChNC surface. Additionally, weak bands observed around 800 cm^-1 can be attributed to overlapping contributions from phosphate groups (P–O bending vibrations) associated with the PBS, as well as C–C and C–O skeletal vibrations from the polysaccharide matrix (Tadda et al., 2021) with possible contributions from protein-related vibrations due to the antibody, although this region lacks specificity.

Quantitative analysis was then done by titrations (Table 1), where the polyelectrolyte titration provided an overall value of charge density. Contributions of the carboxy and amino groups were specifically measured by conductimetric titrations.

The CNCs presented the lowest charge density of all the nanocrystals compared, with a polyelectrolyte consumption of 936.67 ± 42.9 µeq/g, which corresponded to 34.7 ± 1.9 μmol/g of the sulfur ester added during hydrolysis, and 469.3 ± 0.0 μmol/g of the carboxy groups that result from the partial oxidation during the processing. Once the chitosan was adsorbed to the surface, the overall charge increased to 2531.1 ± 36.7 µeq/g, with most of it resulting from the 3.58 ± 0.16 mmol of amino groups added per gram of nanocrystal, as there was no significant change in the carboxy content. The ChNC presented a charge density of 1931.7 ± 48.2 µeq/g, with 0.86 ± 0.11 mmol of amino moieties per gram of material. The great difference between these values arises from the ability of the chitosan amino groups to protonate/deprotonate over a broad pH range and may be partially inaccessible in heterogeneous or crystalline domains of the nanocrystal. Additionally, high ionic strength in PBS can screen electrostatic contributions to ζ-potential, so low ζ-potential values do not necessarily contradict substantial titratable amino content or charge density measured under titration conditions.

Based on the functional group densities reported in Table 1 and the nanocrystal mass used during immobilization, the materials contained a substantial excess of reactive groups relative to the antibody. The antibody dose corresponded to approximately 1.33 pmol of IgG molecules, assuming a representative molecular weight of 150 kDa. Because each IgG contains multiple reactive residues, functional group availability was compared using antibody reactive-group equivalents rather than whole-antibody equivalents. Assuming approximately 90 primary amino groups and 150 carboxy groups per IgG, the estimated material-to-antibody reactive group ratios were ∼1.06 × 10⁵ for CNC carboxy groups relative to antibody amino groups and ∼1.16 × 10⁵ for ChNC amino groups relative to antibody carboxy groups. For CNC/Ch, the corresponding individual ratios were ∼9.30 × 10⁴ for carboxy groups relative to antibody amino groups and ∼4.83 × 10⁵ for amino groups relative to antibody carboxy groups; considering both CNC/Ch functionalities together gave a combined reactive-group excess of ∼3.37 × 10⁵ relative to total antibody reactive-group equivalents. The full calculations are present in the Supplementary Material. These calculations show that antibody immobilization was not limited by the nominal number of available functional groups, and differences among materials are more likely governed by functional group accessibility, local reactivity, hydration, and interfacial organization.

As expected, the three materials had significantly different surface chemistry, with the composite material presenting the highest value as both carboxy and amino groups are present at the surface. After antibody immobilization, the CNCs and CNC/Ch presented a significant decrease in surface charge, further confirming the surface modification. For the ChNC, no significant change in the charge density was measured despite the changes in the FTIR spectra that were measured. This can reflect that EDC/NHS-mediated coupling and protein association change local bonding/hydration without measurably changing the net effective charge under the polyelectrolyte titration conditions.

3.2 Real-time antigen detection

Antigen detection for all the tested materials was first monitored using the QCM-D (Figures 4a–d; Supplementary Figure S1) with PBS used as media to maintain the antigen and antibody integrity and tonic stability. Antigen concentrations were increased in stepwise injections, aiming to identify the antibody saturation plateau. The change resulting from each injection was calculated by subtracting the values obtained after rinsing from the previous PBS baseline. Each nanocrystal was tested with and without antibodies to differentiate contributions from the antibody-antigen conjugation against the electrostatic interactions and hydrophobic associations. In terms of assay performance, QCM-D and MP-SPR provide complementary descriptors: QCM-D signals reflect combined effects of antigen binding, hydration, and viscoelastic effects, which can be associated with apparent sensitivity at low concentrations, whereas MP-SPR is less affected by solvent effects, providing another effective approach to quantify antigen binding. Therefore, responses from both techniques are interpreted as relative indicators of performance under the tested conditions rather than absolute measures of assay sensitivity. Consequently, the experimental protocol was designed to distinguish stable binding from reversible interactions.

For each experiment, the initial baseline region was obtained by flowing PBS over the precoated sensor until a stable baseline was achieved, indicating equilibrium between the sensor surface and the buffer environment. Once stability was reached, the signal was recorded for an additional 30 min. This baseline step also removed any loosely bound material from the sensor surface, ensuring that subsequent responses originated from firmly adhered coatings. Following baseline stabilization, a 10% antigen solution diluted in PBS was introduced and allowed to interact with the surface for 30 min. This procedure was sequentially repeated for the remaining antigen concentrations (50%, 100%, 200%, and 400%). Between each antigen exposure, PBS was flowed over the sensors for 30 min to remove unbound or weakly adsorbed antigen. This rinsing step enabled discrimination between reversible and stable binding events, aiming to ensure that observed changes in frequency and dissipation were primarily associated with the surface interactions occurring at the coating interface and not by multilayer formation. The average values of adsorption and dissipation and their standard deviation are presented fully in the Supplementary Material.

In the absence of antibody functionalization, the three nanocrystals exhibited different QCM-D responses to antigen dilution. CNC films showed minor shifts in ΔF and ΔD, with a clear increase in mass deposition at the highest concentrations, suggesting weak, nonspecific antigen adsorption. The CNC/Ch composite exhibited changes in frequency, dissipation, and Sauerbrey-derived mass in the single digits across all antigen dilutions. Contrary to the pristine CNCs, the adsorption seems to follow a linear increase from the lowest antigen concentration. In contrast, ChNC displayed a markedly stronger response. It showed pronounced negative frequency shifts and large positive mass uptake (Figures 4b,c), particularly at intermediate dilutions (50%–100%), accompanied by significant increases in dissipation. At higher antigen dilutions (200%–400%), the intensity of the changes in frequency and mass decreased, suggesting that at higher antigen concentration, there was a more notorious interplay between the antigen adsorbed and the release of material (water, salts, and antigens), as these changes were accompanied by a decrease in dissipation (the stiffening of the model film). Because ChNC layers are highly hydrated and responsive to ionic screening, the apparent decrease at high concentration may also reflect saturation-driven compaction/deswelling and reduced coupled-water contribution rather than reduced biochemical interaction alone.

Although the extent of the effect depended strongly on the underlying substrate, antibody functionalization significantly altered the QCM-D response of all materials. CNC-Ab exhibited a more pronounced response than CNC alone. From all the materials, CNC-Ab presented the highest response of all the materials, suggesting the best detection at low concentrations. This trend continued throughout all the antigen dilutions in an almost linear way. Dissipation values remained low across all conditions but showed the only significant difference on the pair comparison (p = 0.03) when compared against the pristine CNCs. This difference confirms the effects of the antigens in the film viscoelastic properties and hydration. CNC/Ch-Ab exhibited a constant increase in the adsorbed mass, suggesting that at the beginning, the film structure was changing, liberating water/salt in greater quantities than the antigen being detected. This is also observed in the dissipation values, where an increase in rigidity was observed at the beginning of the measurements and increased slightly at the consequent injections.

The ChNC-Ab showed a significantly different behavior than the ChNC alone (p = 0.0007), reflecting substantial restructuring upon antigen exposure. Immediately after the first antigen injection, a sharp frequency increase (apparent mass loss) was observed. Although less pronounced at higher antigen concentrations, apparent mass loss persisted throughout the experiment, albeit in a non-proportional manner. This initial frequency increase likely reflects expulsion of loosely associated buffer anions (e.g., chloride or phosphate) and hydration water from the protonated chitosan surface upon antigen binding and partial charge screening. The reduced magnitude at higher antigen concentrations is consistent with progressive site saturation and prior film compaction, leaving fewer loosely associated ions and hydration water available for expulsion. From all the pair comparisons, only the one between CNC-Ab and ChNC-Ab was significantly different (p = 0.0058). This can clearly be seen in both the frequency (Figure 4b) and mass (Figure 4d) plots, where the CNC-Ab shows a clear adsorption behavior, while the ChNC-Ab loses mass. Importantly, this “mass loss” is most likely dominated by hydration/ion release and viscoelastic restructuring as QCM-D alone does not directly prove physical detachment of the capture layer.

The contrast in response between the antibody-containing CNC and CNC/Ch surfaces with the ChNC-Ab, suggests that the ChNC-Ab likely formed a hydrated, viscoelastic layer, rather than a rigid Sauerbrey-type layer (Dixon, 2008). Upon antigen binding, charge screening of protonated amino groups likely induced film compaction and partial deswelling, resulting in the release of coupled water and counterions. The observed frequency increases, therefore, reflect structural rearrangement within the hydrated chitosan layer rather than detachment of the capture film. Such behavior is consistent with the polyelectrolyte nature of many natural polymers and previously reported diffusion- and hydration-mediated effects in CNC-Ab immunoassay films (Nori et al., 2023). Because QCM-D is sensitive to coupled water and viscoelastic changes, concurrent alterations in hydration and film mechanics can strongly influence the acoustic response (Su et al., 2005). Therefore, variations in response levels observed across different materials may result not only from antigen binding but also from variations in hydration and film restructuring.

To disambiguate these coupled processes, MP-SPR provides an orthogonal optical readout that is substantially less sensitive to associated water and more directly proportional to the accumulation of interfacial “dry” mass (via refractive-index changes) at the gold surface (Fang et al., 2018). The combined interpretation of acoustic (QCM-D) and optical (MP-SPR) signatures is a well-established strategy for adsorption studies on cellulosic interfaces because it enables separation of concentration-dependent binding from hydration- and mechanics-driven artifacts that can dominate in soft, heterogeneous films (Pękala et al., 2023). In this context, discrepancies between QCM-D and MP-SPR responses are interpreted as technique-dependent rather than indicative of intrinsic material superiority.

The MP-SPR data for each sensogram (representative one in Figure 4e, all in Supplementary Figure S2) was summarized in a similar manner to that of the QCM-D, to compare the irreversible adsorption (Δ°) after each injection (Figure 4f). Overall, the CNC-only surfaces exhibited a wide signal change across repetitions and antigen concentrations, consistent with limited non-specific interaction. Once coated with Ch, the composite (CNC/Ch) showed the smallest angle shift, indicating linearity like that observed in QCM-D at the 1:1 antigen-antibody concentration (100%). Likewise, the ChNC surfaces showed adsorption at all antigen concentrations, likely driven by electrostatic interactions and hydrophobic associations.

Once antibodies were immobilized, the CNC-Ab films showed a narrower adsorption behavior than the pristine CNC films, with a sustained increase in signal at higher concentrations. In contrast, the antibody immobilized onto the composite CNC/Ch (CNC/Ch–Ab) exhibited a more consistent and increased optical response at intermediate-to-high antigen concentrations (100%–400%), indicating enhanced antigen binding under these conditions. From all the statistical comparisons, this was the only pair that had a significant difference (p = 0.02) for the ChNC-Ab films. Although ChNC–Ab exhibited broader variability relative to pristine ChNC, the overall angle shifts were lower, suggesting reduced non-specific adsorption and improved selective interaction profile under the tested conditions.

Discrepancies between QCM-D and MP-SPR responses were observed across all materials and concentration ranges, suggesting strong interfacial behavior differences. These variations are likely influenced by inherent sensitivity of each technique to distinguish physicochemical properties, as well as by experimental conditions.

3.3 Signal analysis of prototype of paper detection device

To translate the adsorption results to colorimetric detection, the tested nanocrystals with antibodies were deposited onto additive-free paper (Whatman 5®). Figure 5a shows the picture of the tested materials. At first glance, the CNC-Ab formed a homogeneous deposition which produced a visible response, like the signal produced by the positive control (CNC/Ch immobilized antigens) and stronger than the negative control (free antibody). The CNC/Ch also formed a visible spot with an apparent greater deposition of the GNP-Ab on its surface; conversely, the ChNC did not adhere to the paper in a homogeneous way, flowing upwards with the flow of the PBS-antigen suspension. This dragging of the ChNC is consistent with observations in the QCM-D, where the Ab-carrying ChNC tended to flow once they interacted with the antigen, supporting the notion that ChNC-based layers are highly hydrated/viscoelastic and can undergo restructuring with coupled water/ion release, which may correlate with weaker retention on porous cellulose substrates. As some adsorption of GNPs to the paper was still occurring and visual changes were only qualitative, the visual intensity was quantified to enable unbiased comparison. Based on the quantified color differences (Section 2.8.3) between the materials (Figure 5b), all of them had values greater than the free antibody (0.37 ± 0.2) used as a negative control. The materials, per se, did not have a significant difference between them, but the CNC/Ch-Ab (3.1 ± 0.2) had a narrower distribution than the CNC-Ab (3.2 ± 0.8) and the ChNC (4.4 ± 0.5).

As the most homogeneous signal was observed with the CNC/Ch-Ab composite, variation in the antigen’s concentration was also tested (Figure 5c). Visible signals were obtained at the three concentrations tested (4, 20, and 60 μg/mL), with no significant difference between them, showing that the material was sensitive enough across concentrations. Moreover, none of the concentrations visibly altered the definition of the testing spot or produced any detachment with drag, like with the ChNC.

4 Discussion

The tested composites (CNC-Ab; ChNC-Ab; CNC/Ch-Ab) demonstrate that antibody immobilization efficiency is not governed by a single physicochemical parameter, but rather by the interplay between surface chemistry, charge density, interfacial organization, and colloidal stability. Previous studies using CNCs derived from wood have demonstrated that the density of sulfate groups directly influences the efficiency of antibody immobilization (Basso et al., 2025; Martínez-Negro et al., 2023); however, such systems often face limitations related to electrostatic repulsion and require intermediate silane layers (e.g., APTES–GA) to ensure film stabilization on the transducer surface (Amit et al., 2024). In contrast, the cotton-derived CNCs used in the present study exhibited a high native carboxy content (469.3 μmol/g), enabling direct activation via EDC/NHS chemistry, whereas the incorporation of chitosan introduced positively charged amino groups known for forming stable electrostatic complexes (3.58 mmol/g). This intrinsic functionality resulted in a significant reduction of surface charge after conjugation and improved colloidal stability (Gu et al., 2017; Zubareva et al., 2016). These findings support the hypothesis that efficient antibody immobilization depends not only on the presence of reactive groups, but on the balance between surface charge type and density. In this context, accessible carboxy groups enable stable covalent attachment, while protonated amino groups contribute to electrostatic stabilization and may promote more favorable interfacial organization of the antibody.

The intrinsic reactivity of the natural nanocrystals’ surfaces simplifies the immobilization strategy by eliminating the need for a silane “bridging” layer that was used with the APTES-CNC (Nori et al., 2023). Furthermore, cotton, composed of approximately 99% cellulose, contains fewer residual lignin or hemicellulose components than wood-derived sources, potentially leading to improved surface homogeneity and interfacial reproducibility. However, the higher charge density of cotton-derived CNCs also introduces challenges. A strong tendency toward aggregation was observed in saline medium (PBS), with hydrodynamic diameters exceeding 1700 nm. Such aggregation reduces the effective surface area exposed to the analyte and introduces variability in QCM-D dissipation signals, particularly in hydrated films. By comparison, wood-derived CNCs, typically shorter (∼165 nm), can form more densely packed and mechanically stable thin films on piezoelectric transducers, enabling extremely high immobilization capacities (∼6300 ng/cm²)(Amit et al., 2024).

The aggregation behavior of cotton-derived CNCs in PBS is consistent with previous reports describing self-association phenomena in phosphate-containing media due to differences in hydrophilicity and polarity between the highly hydrophilic CNC surface and the buffer environment (Al-Nakashli et al., 2023; Kwak et al., 2023). This interfacial incompatibility promotes particle–particle interactions and limits the stabilizing effects of electrostatic repulsion, favoring aggregate formation. Strategies such as surface modification with PBS-compatible oligomers (e.g., succinic anhydride and 1,4-butanediol derivatives) have been proposed to improve colloidal dispersibility (Kwak et al., 2023; Nori et al., 2023; Reid et al., 2026). In the present study, however, despite hydrodynamic aggregation, the functional response was not compromised. The formation of the CNC/Ch composite proved decisive in balancing these effects. The introduction of chitosan added primary amino groups (3.58 ± 0.16 mmol/g), substantially increasing the total charge density (2531.1 ± 36.7 µeq/g) and creating a dual-functional interface of carboxy and amino groups.

These results suggest that surface area alone does not govern sensing performance; rather, effective binding arises from the interplay between surface chemistry, interfacial organization, and film stability. While nanocrystals are often assumed to benefit from increased areas, the present results indicate that chemical functionality and stability of the interfacial layer are critical for analytical response. Although ChNCs have a smaller hydrodynamic size (∼293 nm) and theoretically higher surface area, they did not yield the best analytical response, likely due to instability of the hydrated interfacial layer. In contrast, the CNC/Ch composite combines high surface area with dual chemical functionality (carboxy and amino groups) (Golmohammadi et al., 2017). This dual chemistry provides multiple anchoring mechanisms via EDC/NHS coupling and electrostatic stabilization (Villa et al., 2025). Consequently, the composite enhances effective binding rather than simply increasing the number of available sites. This indicates that a high surface area is insufficient if the interfacial layer lacks structural organization, as seen in systems where excessive loading leads to steric hindrance and decreased sensitivity. Optimal performance arises from the balance between chemical functionality and a controlled interfacial structure.

In addition, the stability and orientation of antibodies on the different substrates are also critical factors for the diagnostic applicability of the system. On ChNC surfaces, the high degree of hydration and interfacial fluctuations suggest a less stable immobilization, which may result in random orientation or partial loss of functional conformation of the protein. In contrast, the CNC/Ch composite matrix provides a structural framework that limits excessive swelling of the chitosan network, likely promoting a more favorable antibody orientation. In this configuration, the Fc region is preferentially associated with the bifunctional interface, maintaining the regions accessible for antigen recognition. Such controlled interfacial organization not only preserves the biological functionality of the antibody but also enhances detection reliability by minimizing steric hindrance effects, which are commonly observed in systems with heterogeneous surface coverage.

The composites also revealed significant differences in interfacial behavior monitored by QCM-D and MP-SPR. Previous studies on wood-derived CNC films, particularly after APTES–GA functionalization, have reported relatively rigid-layer behavior compatible with the direct application of the Sauerbrey equation (Amit et al., 2024). In contrast, cotton-derived CNCs and especially the CNC/Ch composite exhibited a noticeable viscoelastic and hydration contribution. CNC-Ab films displayed progressive antigen adsorption with low dissipation, indicating the formation of a relatively rigid layer. In contrast, ChNC-Ab films showed apparent mass loss accompanied by increased dissipation, a response consistent with structural reorganization and coupled water release rather than film detachment.

The CNC/Ch-Ab system exhibited intermediate and more favorable behavior. An initial interfacial reorganization was observed, followed by mechanical stabilization and a consistent increase in detected mass in both QCM-D and MP-SPR measurements. Importantly, MP-SPR is less sensitive to coupled water. The optical results confirmed that the observed signal increase corresponded to real antigen accumulation rather than mechanical artifacts. These results reinforce that CNC source influences not only surface chemistry but also the mechanical properties of the film formed on the sensor, directly affecting data interpretation. In this context, material performance is dependent on both the measurement technique and antigen concentration range: while CNC-Ab shows stronger responses at low concentrations in QCM-D, CNC/Ch–Ab exhibits more consistent behavior at higher concentrations in MP-SPR, indicating that the apparent advantages of each system are condition-dependent. Although ChNCs exhibit a smaller hydrodynamic diameter (∼293 nm), this did not translate into superior sensing performance. In label-free techniques like QCM-D and SPR, the response is heavily dictated by the physicochemical state of the interface, including hydration, viscoelasticity, and matrix reorganization (Reviakine et al., 2011). Chitosan acts as a weak polycationic polyelectrolyte (pK _a ≈ 6.5) due to its high density of primary amino groups. Under buffered conditions, ionic strength and pH influence the degree of protonation and electrostatic screening, which in turn affect chain conformation, swelling behavior, and interfacial stability. These effects can lead to increased hydration and structural fluctuations in ChNC-based systems, contributing to less stable and less reproducible sensing responses.

In physiological buffers such as 1x PBS (≈0.15 M ionic strength) (St-Pierre and Sato, 2022), electrostatic interactions are significantly screened. This phenomenon reduces repulsion between protonated amino groups, triggering a conformational rearrangement of the polymer network. Under these conditions, chitosan interfaces form diffuse, highly hydrated layers. These layers are sensitive to ionic strength and undergo structural transitions often characterized by slow equilibration and hysteresis (Hollingsworth and Larson, 2021; Smook et al., 2025). As a result, the interfacial layer exists in a dynamic equilibrium between swollen and partially collapsed states. This behavior is well-documented for polyelectrolyte films, which undergo reversible swelling/deswelling and counterion-dependent restructuring (Miras et al., 2021). QCM-D studies demonstrate that frequency shifts (Δf) and dissipation changes (ΔD) are linked to variations in film thickness and viscoelasticity (Miras et al., 2021). These structural alterations can have a magnitude comparable to actual biomolecular binding signals. Therefore, the sensor output reflects not only mass adsorption but also the evolving physicochemical state of the interface. During antigen exposure, the chitosan coating contributes to matrix reorganization, including water release and compaction. These coupled processes can lead to signal instability and reduced reproducibility in QCM-D and SPR measurements. Consequently, a high nominal surface area does not guarantee better detection if the interfacial layer lacks structural rigidity. This interpretation is consistent with Figure 4 data, where the ChNC-Ab system exhibited higher dissipation and inconsistent acoustic responses. In contrast, the CNC/Ch-Ab composite showed a consistent decrease in ΔF/n with moderate increases in ΔD. This suggests that the composite promotes stable mass accumulation within a hydrated, yet structurally coherent and stable matrix.

Conversely, on the paper-based prototype, reproducibility was confirmed by the consistency of the colorimetric signals. The CNC/Ch-Ab composite exhibited the lowest variability (3.1 ± 0.2), which is attributed to the formation of a structurally coherent matrix that stabilizes the detection interface. In contrast, the ChNC-Ab system showed higher variability (4.4 ± 0.5) and signal instability (Figure 5). This behavior is directly correlated with the high degree of hydration and swelling of pure chitosan layers. In buffered media, these polyelectrolyte interfaces become highly viscous and diffuse, undergoing structural reorganization and release of coupled water during interaction with the analyte (Reviakine et al., 2011). This mechanical instability, driven by excessive swelling, facilitates partial detachment or redistribution of the material during flow along the paper substrate, resulting in less reproducible responses. The synergistic effect in the CNC/Ch composite mitigates these hydration-related effects by limiting network swelling and ensuring that chemical immobilization prevails over physical fluctuations of the interface. Such adjustments in interfacial organization are critical, as recent studies have demonstrated that surface coverage uniformity is more important for detection efficiency than the nominal number of immobilized antibodies (Villa et al., 2025).

The high variability observed in the CNC-Ab and ChNC-Ab samples after deposition further confirms that hydration and swelling of the interfacial detection layer are critical factors that must be addressed (Golmohammadi et al., 2017; Liana et al., 2012). Traditionally, protein-based sensors are developed using nitrocellulose as the substrate, as it offers low cost, paper-like behavior, and surface hydrophobicity (Holstein et al., 2015; Liana et al., 2012). However, this material lacks long-term chemical stability and hinders biodegradability. Glossy paper has emerged as a valuable alternative, demonstrating good performance (Liana et al., 2012); nevertheless, since cost and widespread availability are essential for point-of-care applications, the use of generic paper remains the most practical option (Syedmoradi et al., 2017). Insights from other fields, such as packaging, can provide guidance on how to easily coat paper with biopolymers to enhance surface stability, modulate capillarity, and enable homogeneous deposition of active nanocrystals (Herrera et al., 2025; Rastogi and Samyn, 2015).

Regarding analytical sensitivity, when compared to previous studies employing sulfated wood-derived CNCs for CEA detection, (Nori et al., 2023), the present system demonstrated measurable antigen response starting from the lowest concentration tested (80 ng/mL). Although this concentration range falls within values commonly employed in nanocellulose-based QCM studies, the progressive and reproducible signal observed across increasing antigen levels confirms efficient interfacial recognition without the need for intermediate amplification layers. Notably, the CNC/Ch-Ab platform generated coherent responses at low concentrations while maintaining controlled viscoelastic behavior, indicating that the observed sensitivity is not merely a consequence of excessive film swelling or nonspecific adsorption. While the present work does not aim to establish an ultralow limit of detection, it provides robust evidence that structurally optimized nanocrystal interfaces can achieve reliable antigen detection within clinically relevant concentration ranges, with improved interfacial stability compared to previously reported systems.

Despite the promising results, some limitations of this study should be acknowledged. In particular, the orientation and conformational state of the immobilized antibodies were not directly quantified and were instead inferred from interfacial behavior and sensing performance. Future work should focus on developing strategies to explicitly characterize antibody orientation at the interface, which would provide a more rigorous understanding of structure–function relationships. In addition, evaluating different types of proteins with varying sizes and binding characteristics would help reduce interpretation bias and confirm the general applicability of the proposed platform. Likewise, the limit of detection and other factors, including complex biological matrices, still need to be studied to demonstrate the capability of this system in real-world scenarios. At present, this preliminary work demonstrates a promising trend in the use of natural polymers as analytical tools and provides a strong rationale for future studies for the incremental understanding and robustness of the system.

5 Conclusion

Through the comparison of the different renewable nanomaterials, this work demonstrates that the immobilization of antibodies is highly driven by the interfacial interactions in suspension, and that the surface chemistry of the materials can influence but not determine sensitivity. A critical finding was the great effect that the swelling and hydration of the detection layer has on the detection capability of the nanocrystals, which, due to their polysaccharide origin, tend to be hydrophilic. The formation of a composite material, CNC and chitosan, provided stability and lower variability in the detection, seemingly providing the best of both natural polymers. The prototype of a paper-based detection device demonstrated high sensing density, yielding measurable signals even at concentrations lower than those used in specialized tests such as ELISA. Nevertheless, optimization of the paper substrate is needed to increase contrast and to prevent the dragging of the nanocrystals due to the wicking of the buffer containing the antigen. Moreover, testing with simulated human fluids and real-life samples still needs to be performed, as the protein content of those might change the sensitivity and selectivity of both the analytical probe and the paper substrate. Overall, the results of this study offer a strong foundation for the development of renewable polysaccharide-based nanomaterials as immunosensors for paper-based detection.

Generative AI statement

The author(s) declared that generative AI was used in the creation of this manuscript. During the preparation of this work, the author(s) used ChatGPT v.5.2 to correct grammatical errors and improve readability. After using this tool/service, the author(s) reviewed and edited the content as needed and take(s) full responsibility for the content of the published article.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fbiom.2026.1815047/full#supplementary-material

Supplementary material Supplementary material at the source

Footnotes

Edited by: Antonio Topete, University of Santiago de Compostela, Spain

Reviewed by: Juan M. Ruso, University of Santiago de Compostela, Spain

Juliane Glória, Instituto Leônidas and Maria Deane (ILMD/Fiocruz Amazônia), Brazil