3.1. Sensor Characterization
We employed the INPS strategy to monitor HSA adsorption and denaturation on dielectric-coated sensor arrays, which were comprised of randomly-distributed gold nanodisks on glass surfaces fabricated via hole-mask colloidal lithography and the entire surface was then conformally coated with a 10-nm thick silica or titania overlayer (
Figure 1A). Briefly, when HSA solution is introduced into the microfluidic chamber, HSA molecules adsorb onto the sensor surface and accumulate to gradually form a dense protein adlayer, as shown in
Figure 1B. The entire process occurs under the illumination of light, which passes through the sensor to reach the spectrophotometer. The adsorption of HSA molecules onto the sensor surface leads to a Δλ
max shift, and our measurement approach is sensitive to the dry mass of adsorbed HSA molecules.
The optical extinction spectra corresponding to before and after HSA adsorption onto silica- and titania-coated gold nanodisks are shown in
Figure 2A. The presence of different dielectric coatings leads to a significant variation in the initial λ
max position (i.e., ~715 nm and ~767 nm for silica- and titania-coated sensors, respectively), which arises from perturbations in the electromagnetic field distribution around the gold nanodisks due to the presence of the dielectric coatings [
64,
65,
66]. Consequently, the distinct plasmonic properties result in the sensor arrays exhibiting different bulk refractive index sensitivities, which typically increase as the LSPR peak position moves to higher wavelengths [
67,
68]. We determined the bulk refractive index sensitivities to be ~100 and ~140 nm/RIU for bare silica- and titania-coated sensors, respectively (
Figure 2B). While electromagnetic field enhancement around the edges of bare plasmonic nanostructures is often amplified by up to several orders of magnitude more than flat or smooth regions [
69,
70,
71], past simulation results suggest a relatively uniform electromagnetic field distribution around the contacting surface of dielectric-coated gold nanodisks [
11].
After HSA adsorption under physiologically-relevant ionic strength conditions (i.e., 150 mM NaCl) for 20 min, a positive Δλ
max shift was observed on both surfaces. While the magnitude of the Δλ
max shift (~1 nm) is relatively small, it is significantly larger than the sensing resolution by several orders of magnitude. This was verified by directly characterizing the sensing resolution, which was defined as the smallest change in the bulk refractive index that produced a detectable change in the output [
72]. The practical resolutions were determined to be 9.20 × 10
−5 RIU and 1.12 × 10
−4 RIU, which translates to Δλ
max shifts of around 0.009 nm and 0.016 nm, for silica- and titania-coated sensors, respectively (details in
Supplementary Materials). Taken together, these findings identify that the fabricated titania-coated arrays have ~1.4-times greater measurement sensitivity than the silica-coated arrays, and establish that HSA protein molecules adsorb onto both sensor surfaces producing detectable measurement responses.
3.2. Real-Time Monitoring of HSA Adsorption Kinetics
To resolve protein adsorption kinetics, we tracked the Δλ
max shift as a function of time. As shown in
Figure 3A, the time-resolved peak shifts indicated that HSA adsorption onto silica and titania surfaces shows similar adsorption profiles under physiologically relevant ionic strength conditions (i.e., 150 mM NaCl). In particular, there were similar adsorption rates in the initial and equilibration stages, while one distinct difference was a significantly higher final peak shift for the titania case. The final peak shifts were around 0.9 nm and 1.4 nm for HSA adsorption onto silica and titania, respectively, and the range of these values agrees well with literature values for BSA adsorption [
37,
38,
43], validating our measurement approach. The LSPR signals can be normalized by the bulk refractive index sensitivities to directly compare the measurement responses across different substrates [
11]. After normalizing the measurement responses by the bulk refractive index sensitivities, it is noticeable that the extent of HSA adsorption was nearly identical on the two substrates (
Figure 3B). Specifically, the rates in the initial and equilibration stages were slightly higher for HSA adsorption onto silica than titania surfaces, with the trend switching in the intermediate stage.
It is known that the presence of salt can affect protein-protein and protein-substrate interactions through electrostatic shielding effects [
56,
73]. In addition, protein conformation can also be stabilized by the presence of salt [
74,
75,
76,
77]. Since a silica surface is more negatively charged than a titania surface under the tested pH conditions [
52], we proceeded to investigate HSA adsorption onto both surfaces under different ionic strength conditions ranging from 0 mM to 250 mM NaCl to evaluate surface charge effects. Interestingly, significant differences in adsorption behavior became apparent at different ionic strengths. With increasing ionic strength, there was a decrease in the normalized peak shifts for the case of HSA adsorption onto silica, indicating lower accumulation of protein molecules within the immediate vicinity of the gold nanodisks (
Figure 4A). In marked contrast to the silica case, minimal variations in the measurement response were observed for the titania case (
Figure 4B).
In order to verify the observed measurement trends, we performed quartz crystal microbalance-dissipation (QCM-D) experiments using silica- and titania-coated quartz crystals. The QCM-D technique tracks changes in the resonance frequency and energy dissipation of an oscillating quartz crystal that reflect the acoustic mass and viscoelastic properties of an adsorbate, respectively. With increasing ionic strength, there was an overall decrease in the QCM-D frequency shift for the case of HSA adsorption onto silica, indicating lower accumulation of protein molecules on the surface (
Figure 5A). Compared to the LSPR peak shifts, the QCM-D frequency and dissipation shifts for HSA adsorption onto silica were less varied across the tested range of ionic strength conditions. In contrast, frequency shifts closely overlapped for the case of HSA adsorption onto titania, except for the case of 0 mM NaCl, which indicated the lowest amount of adsorbed HSA (
Figure 5B). The accompanying energy dissipation shifts showed similar values during the stabilization stage for HSA adsorption onto silica, suggesting that protein molecules are closely packed to form a compact film and the specific configuration likely depends on the particular ionic strength condition (
Figure 5C). Likewise, energy dissipation shifts for HSA adsorption onto titania at different ionic strengths were also similar, although the values were much lower compared to the case of silica, indicating that HSA molecules are more deformed and tightly bound on the titania surface (
Figure 5D). Overall, the trends observed via QCM-D measurements were in excellent agreement with those observed via LSPR measurements.
3.3. Data Analysis
A detailed analysis of the LSPR response requires some prior knowledge about the biomacromolecular interaction under consideration. In general, a protein molecule initially adsorbs onto a surface in its native conformation and then, depending on the protein-surface interaction, either remains in its native conformation or unfolds and spreads on the surface [
42,
78,
79,
80]. Upon spreading, the surface area covered by a protein molecule, as well as the binding strength between the surface and protein molecule increases [
81,
82]. Depending on the extent of protein deformation, protein molecules can either remain adsorbed or desorb from the surface. Over time, the protein adsorption process will become less reversible and the extent of protein spreading will increase after multiple adsorption-desorption cycles [
82,
83,
84,
85,
86].
When interpreting LSPR measurement responses, it is often assumed that protein density remains constant [
87,
88] leading to a direct correlation between LSPR peak shift and optical mass. However, as mentioned above, both protein density and thickness of the protein film change. As described in our previous works [
11,
13,
14], the peak shift,
, arising from a refractive index change in the LSPR-probed sensing volume around the nanodisk is denoted by:
where
is the bulk refractive index sensitivity of the sensing arrays (i.e., 100 nm/RIU and 140 nm/RIU for silica- and titania-coated surfaces, respectively),
is the coordinate perpendicular to the substrate surface (
corresponds to the protein-substrate),
is the length scale characterizing the distance between the center of the nanodisk and the protein-substrate contact (i.e., 74 nm as previously determined by simulation results [
11]), and
is the spatial distribution of the refractive index change along the
coordinate. Considering the formation of a protein layer directly on the sensor surface, integration of Equation (1) yields:
where
is the normalized peak shift and
is the refractive index change arising from the adsorption of protein molecules to form a protein layer with thickness
. We applied the above correlation to approximate the thickness of the HSA layer based on typical values of local refractive index changes due to HSA adsorption onto titania, which is in the range of 0.08 to 0.12 RIU after 30 mins [
88]. Rearranging Equation (2), we obtain:
Substituting the typical values of
, as well as
from
Figure 4A, B into Equation (3), we obtain a protein film thickness of around 1.5 nm for adsorbed protein layers on both surfaces at physiological ionic strength condition, which agrees well with previous works [
38,
39]. Based on the above arguments, the LSPR measurement response (Δλ
max shift) can arise from either protein adsorption (mass accumulation) on the surface and/or from substrate-induced conformational changes [
43,
89]. Applying this framework to protein films requires a delicate approach because (1) the majority of albumin molecules will desorb shortly after they initially adsorb on the surface, especially when there is low surface coverage [
90]. Over time, (2) HSA molecules that remain irreversibly adsorbed reduce the available surface area for subsequent adsorption events and, as a consequence, (3) there will gradually be a lower flux of newly adsorbing proteins [
82,
86].
Having established the framework for interpreting LSPR responses, let us first analyze the data for HSA adsorption on silica and titania at physiological ionic strength. By comparing the time-resolved peak shifts in
Figure 3B, we can deduce that the initial rate of reversible adsorption was higher on silica, while the tendency for protein spreading is lower. Consequently, protein spreading on silica occurred more gradually over a longer time period. On the other hand, a quicker and higher degree of HSA spreading occurred on titania. The evidence supports that, due to a higher surface charge, the initial adsorption of HSA on silica is more strongly governed by long-range electrostatic interactions [
91,
92], compared to that on titania where the adsorption process is less driven by electrostatic forces (i.e., no dependence on ionic strength) [
93,
94]. Conversely, this supports that HSA spreading on titania occurs due to stronger, short-range protein-surface interactions stemming from non-electrostatic forces (e.g., van der Waals). Furthermore, the LSPR measurements also showed a lower propensity for HSA desorption from titania upon rinsing with buffer (
Figure S1). Of note, Givens et al. recently showed that BSA adsorption onto oxide nanoparticles follows a similar trend, whereby adsorbed BSA is more completely denatured on titania nanoparticle surfaces, while it is less denatured on silica nanoparticle surfaces [
95].
Furthermore, the decrease in initial rate and equilibration uptake with increasing ionic strength on silica suggested lower rates of reversible adsorption and greater spreading. Interestingly, for 0 mM and 50 mM NaCl conditions, the initial rates were closely matching before the rate decreased more considerably for 50 mM NaCl during the intermediate stage. This result is reasonable since it is known that proteins are destabilized in the absence of salt due to the lack of intramolecular charge shielding resulting in greater intramolecular repulsions [
74,
75,
77]. The observations also suggest that HSA spreading and/or reorientation in the later stage was induced by further protein uptake, which did not occur for 50 mM NaCl since the adsorbed HSA molecules were more stable and, therefore, had lesser tendency to reorient to accommodate later arriving molecules [
96]. In contrast, HSA adsorption onto titania did not change significantly as compared to adsorption under physiological ionic strength, supporting that HSA adsorption onto titania is strongly governed by short-range interaction forces. Compared to HSA desorption from silica, HSA desorption from titania was also lower at all tested ionic strengths (
Figure S2). In addition, stronger van der Waals forces on titania also promote the formation of a thicker hydration layer [
97], which may mediate the propensity for protein deformation on the surface, resulting in similar rates of conformational change on the surface independent of ionic strength.
Taken together, our results reveal that, while HSA adsorption onto two hydrophilic surfaces appears to be very similar under physiologically relevant ionic strength conditions, significant variations in adsorption behavior are observed under different ionic strength conditions and can be detected by LSPR measurements. Comparison between experimentally-measured HSA adsorption behaviors on silica and titania surfaces is briefly summarized in
Table 1, and provides insight into the role of different factors, as well as benefits of the INPS sensing approach for measurement discrimination.