2.1.

Polarization Sensitive OCT System

To track both translational and rotational diffusions, a key component of the experimental setup is the polarization sensitive signal collection of the OCT system. A detailed description of our custom polarization sensitive spectral domain OCT system can be found in the previous literature,³⁴ but here we provide a brief overview. The system utilizes an 800 nm center wavelength Ti:Sapphire laser (Griffin; KM Labs) with a 120 nm bandwidth as its light source, as shown in Fig. 1. The incident beam on the sample is horizontally polarized, whereas the reference beam as it returns from the retroreflector is linearly polarized at 45 deg (equal parts horizontal and vertical components) after double pass through a quarter wave plate. Light scattered from the sample interferes with the reference beam and the resulting co-polarized (HH: horizontal in and horizontal out) and cross-polarized (HV: horizontal in and vertical out) components are separated by a polarizing beam splitter and directed to a custom spectrometer. The spectral components of both polarization channels are dispersed by a diffraction grating and simultaneously imaged onto each half of a 4096-pixel line scan camera. The OCT system has an axial resolution of $3\mu \mathrm{m}$ and a transverse resolution of $12\mu \mathrm{m}$ in air, and the power directed on the samples was $\sim 3.0\mathrm{mW}$. The depth of focus is $283\mu \mathrm{m}$. A-line rates were either 25 or 62.5 kHz as indicated in each experiment below.

Fig. 1

PS-OCT system schematic.

2.2.

Diffusion Coefficient Calculations

As a brief review of DS-OCT analysis methods, as previously described in Ref. 27, DS-OCT consists of collecting M-mode OCT images of a GNR-laden sample to depth-resolve ($z$) translational ($D_T$) and rotational ($D_R$) diffusion coefficients. M-mode OCT images can then be collected at varying lateral positions ($x$) or at successive times ($t$) to construct a cross-sectional ($x-z$) or dynamic ($t-z$) image of $D_T$ and $D_R$; in this study, we perform cross-sectional imaging on stationary samples (PEO, mucus) and dynamic imaging on the mucus layer of in vitro ALI hBEC cultures.

For each M-mode OCT image, comprised of A-lines sampled at times $t_s$, the co-(HH) and cross-(HV) polarized complex analytic OCT signals, ${\tilde{S}}_{\mathrm{HH}}(t_s;z)$ and ${\tilde{S}}_{\mathrm{HV}}(t_s;z)$, respectively, are computed from raw OCT data by standard methods. At each $z$ position, the time averages are subtracted, then the normalized temporal autocorrelations $(g_{\mathrm{HH}}(\tau ;z),g_{\mathrm{HV}}(\tau ;z))$ are computed. To suppress noise, the autocorrelations are then averaged in depth windows of 20 pixels or less; each window is considered a region of interest (ROI). By the Siegert relation, these second-order autocorrelations are equated to first-order autocorrelations $g^{(1)}(\tau ;z)$, which, for GNRs at their longitudinal resonance, are related to $D_T$ and $D_R$ as follows:³⁵

Under our experimental conditions, the $1/e$ decay time associated with $D_T$ is much longer than that associated with $D_R$, which simplifies Eq. (2) to isolate the $D_R$ term as

Decay times ${\tau }_{\mathrm{ISO}}$ and ${\tau }_{\mathrm{HV}}$ are extracted by fitting Eq. (3) for $\tau \le {\tau }_{1/e}$ and Eq. (4) for $\tau \le {\tau }_{1/e^2}$, where ${\tau }_{1/e}$ and $\tau \le {\tau }_{1/e^2}$ are the delay times at which $g^{(1)}$ falls below $1/e$ and $1/e^2$, respectively; this excludes noisy tails in the autocorrelations from the analysis. One exception to this rule was PEO solutions with concentrations $>1.5\%$, for which a fit range for Eq. (4) only up to ${\tau }_{1/e}$ was used. For in vitro hBEC cultures, a dynamic ${\tau }_{\mathrm{ISO}}$ threshold was applied such that a fit range of $\tau \le {\tau }_{1/e^2}$ was used if fittings had fewer than 4 data points. For all fittings, if fewer than 4 data points were available, the ROI was excluded from further analysis. Also the zero-delay point $g(\tau =0)$ was excluded from all fittings to avoid the contribution of shot noise.

Several criteria were used to determine which ROIs were valid for analysis. Criterion 1 is a top surface detection method to exclude ROIs above the sample or too close to the top surface, which exhibits a strong specular reflection. For PEO and stationary mucus samples, valid ROIs start 15 pixels below the top surface detected ($1\text{pixel}\approx 1.5\mu \mathrm{m}$ in depth) and 5 pixels below for the hBEC cultures. ROIs are 20 pixels in depth for stationary mucus and 3 pixels in depth for hBEC cultures. Criterion 2 excludes ROIs that are below an average co-polarization (HH) signal threshold above the background noise. Criterion 3 excludes ROIs that do not exhibit valid decay times discussed in Ref. 36. Briefly, we reject ROIs where the decay anisotropy ${\tau }_{\mathrm{ISO}}/{\tau }_{\mathrm{HV}}<1.5$, because dynamic scattering from GNRs exhibits higher decay anisotropy. Additionally, ${\tau }_{\mathrm{ISO}}$ and ${\tau }_{\mathrm{HV}}$ must lie within the dynamic range of the OCT system (significantly longer than the sampling time and shorter than the total acquisition time). Criterion 4 applies a threshold for autocorrelation fits based on the coefficient of determination ($R^2$), for PEO solutions $R^2>0.88$ and for stationary mucus $R^2>0.90$. Criterion 5 was applied only to PEO and stationary mucus samples, where all diffusion coefficients more than five standard deviations from the mean were assessed visually for air bubbles and other imaging artifacts and manually excluded if appropriate. The fourth and fifth criteria were not applied to in vitro hBEC culture studies because mucus hydration was expected to be heterogeneous and dynamically changing.

2.3.

Gold Nanorods

GNRs stabilized by cetyltrimethylammonium bromide were synthesized following an established protocol.³⁸ The GNRs were coated with a 2 kDa molecular weight polyethylene glycol (PEG) methyl ether thiol (Sigma Aldrich) to prevent them from adhering to polymer and mucin macromolecules.³⁹ Three batches of GNRs were used in the experiments reported. Transmission electron microscopy (TEM) was performed on each batch to measure the gold core size distributions and PEG coating thicknesses. The mean core sizes were $68\times 19\mathrm{nm}$ for batch 1, $70\times 22\mathrm{nm}$ for batch 2, and $80\times 22\mathrm{nm}$ for batch 3. The thickness of the PEG coating has been measured previously as $\sim 0.5\mathrm{nm}$, which is half of the distance between the sides of nanorods that have dried aligned next to each other on the TEM substrate.²⁷ The sizes of the GNRs were chosen to be on the same scale as the nanopores of mucus ($\sim 0.2$ to $1\mu \mathrm{m}$⁴⁰). Therefore, their Brownian motion is expected to be hindered by the mucus macromolecules, and corresponding diffusion rates are expected to depend on the mucus concentration. The hydrodynamic radius ($R_H$) of the batches, $\sim \mathrm{19,22}$, and 24 nm, respectively, is calculated using GNR length ($L$) and width ($W$) according to the following equation:²⁷

2.4.

Sample Preparations and Experimental Procedures

3.1.

Diffusion Coefficients of GNRs in Polyethylene Oxide Solutions

The first set of experiments involved measuring GNR diffusion rates in PEO solutions because they are easily accessible, and at Megadalton molecular weights their nanostructure can closely resemble that of mucus.³⁷ Figure 2 summarizes these measurements, which were spatially resolved within $x-z$ cross sections of PEO solutions of varying concentrations, then averaged as described above to provide a single value of $D_T$ and $D_R$ for each concentration. The top panel of Fig. 2(a) illustrates the concentration-dependence of the $D_T$ of GNRs in 4 MDa PEO, which exhibits a similar trend as in our previous report using 1 MDa PEO samples with differently sized GNRs:²⁷ $D_T$ in solvent (distilled water) is $\sim 8.5\mu {\mathrm{m}}^2/\mathrm{s}$ and decreases rapidly as concentration increases. At concentrations higher than 1 wt. %, $D_T$ slows down to $<1\mu {\mathrm{m}}^2/\mathrm{s}$. The middle panel of Fig. 2(a) shows the rotational diffusion rate of GNRs measured within the same samples as translational diffusion. The concentration dependence of $D_R$ in 4 MDa PEO exhibits the same trend observed in $D_T$ such that diffusion is hindered by increasing concentration. However, in contrast with the $D_T$ results, the decay of $D_R$ with concentration is less rapid, and $D_R$ continues to decrease across all higher concentrations. $D_R$ is $>5000{\mathrm{rad}}^2/\mathrm{s}$ in solvent and is reduced by a factor of 10 at our highest sample concentration reported (3 wt. %). As shown in the lower panel of Fig. 2(a), in lower concentration samples ($<1$ wt. %), the variability (or precision) of $D_T$ and $D_R$ is $<10\%$. The intrasample variability is determined by dividing the standard deviation by the mean and expressing the result as a percentage, which is often referred to as the coefficient of variation. At higher concentrations, the variability of the $D_T$ measurements increases well over 10%, whereas $D_R$ measurements maintain low variability across all samples. Figures 2(b) and 2(c) display examples of spatially resolved diffusion measurements at low and high concentrations of PEO. Both $D_T$ and $D_R$ diffusion maps appear to be homogeneous with no spatially varying noise, which is consistent with PEO samples being well-mixed. However, in the higher concentration sample of Fig. 2(c), $D_R$ is defined in nearly twice the number of ROIs defined by $D_T$.

Fig. 2

GNR diffusion in aqueous 4 MDa PEO solutions. In the top panels of (a), $D_T$ and $D_R$ are plotted as a function of PEO solids concentration. In many cases, error bars are not visible due to small standard deviation size. Zero concentration represents GNR diffusion in solvent (distilled water). In the bottom panel, intrasample variability is presented as the percent ratio of the standard deviation to the mean for both $D_T$ and $D_R$. In panels (b) and (c), co-polarized (HH) and cross-polarized (HV) B-mode images of the 0.25 and 2.5 wt. % PEO solutions are displayed along with corresponding $D_T$ and $D_R$ values within each ROI derived from B + M-mode images; excluded ROIs based on thresholding criteria described in the methods are displayed as the background color.

The magnitude of the reduction in GNR diffusion in PEO solutions compared to solvent depends on the degree to which the GNRs are confined by the PEO meshwork. The relative degree of confinement may be thought of in terms the size of the hydrodynamic radius of the GNRs, $R_H$, compared to the correlation length of the polymer solution $\xi$. In the prior work, we defined a weakly constrained regime as one where $D_T$ was reduced by less than a factor of 10 from that of solvent and found that this occurred when $R_H/\xi <2.2$ in 1 MDa PEO Ref. 27. Experiments in 4 MDa PEO, we observe a reduction in $D_T$ by a factor of 10 at PEO concentrations $\ge 1.1$ wt. %, which, given an $R_H$ of 19 nm for this nanorods batch and $\xi$ of 14 nm for 4 MDa PEO at 1 wt. %, corresponds to a weakly constrained regime defined by $R_H/\xi <1.4$. For concentrations above 1.1 wt. %, which we will consider to be strongly constrained, $D_T$ was no longer monotonically decreasing and could not be extracted at the highest concentration of 3.0 wt. %. In comparison, $D_R$ continued to monotonically decrease for increasing concentration in the strongly constrained regime, which effectively extends the dynamic range of measurable PEO concentrations to 3.0 wt. % and possibly further. Added support for the ability to extend the dynamic range is seen by assessing the intrasample variability, where fine $D_T$ and $D_R$ precision is evident in the weakly constrained regime, and only $D_R$ precision is fine in the strongly constrained regime.

There are several potential explanations for the varying responses of $D_T$ and $D_R$ versus PEO concentration. First, $D_T$ measurements tended to exhibit lower $R^2$ values from autocorrelation fittings; this caused many ROIs to be rejected via criterion 4 above where the $R^2$ threshold was applied, as observed in Fig. 2(c) at 2.5 wt. %; it is also the reason $D_T$ could not be extracted at the highest concentration measured (3.0 wt. %). We believe that the lower $R^2$ for $D_T$ is either due to the longer ${\tau }_{\mathrm{ISO}}$ values at high concentrations, which become a significant fraction of the measurement time, causing noise artifacts in the autocorrelation traces, or possibly because the strongly constrained GNR motion no longer fits a model that can be described by simple diffusion. In comparison, $D_R$ depends only upon ${\tau }_{\mathrm{HV}}$, which is always shorter than ${\tau }_{\mathrm{ISO}}$ for the GNRs in our experiments due to decay anisotropy;³⁶ as GNR confinement increases and decay times increase, the smaller ${\tau }_{\mathrm{HV}}$ better remains within the dynamic range of the measurement compared to ${\tau }_{\mathrm{ISO}}$. The fact that $D_R$ is more easily measurable may also suggest a fundamental difference in how diffusive prolate particles rotate versus translate during strong confinement in polymeric solutions, an area of active research.⁴⁴

3.2.

Diffusion Coefficients of GNRs in Stationary Pulmonary Mucus

In our previous publications,^27,35 we reported that $D_T$ is sensitive to hBE mucus concentrations $\le 2.5$ wt. %, both stationary and on actively transporting hBE cultures (in vitro). Mucus $\le 2$ wt. % is considered well-hydrated and would be expected during a state of “healthy” pulmonary mucus production. To extend the capability of DS-OCT to measure the lower mucus hydration levels associated with pulmonary disease-like states, we must perform diffusion experiments on more heavily dehydrated mucus samples ($>3$ wt. %). Here we expand our investigation of GNR diffusion to a broader range of mucus concentration, up to 6.4 wt. %, and report values of $D_R$ in hBE mucus for the first time.

As shown in Fig. 3(a), while trial 2 tended to exhibit higher $D_T$ and $D_R$ values at all mucus concentrations relative to trial 1, similar concentration-dependent trends were observed overall. In mucus samples below 2 wt. %, $D_T$ quickly decreases with respect to concentration while remaining above $2\mu {\mathrm{m}}^2/\mathrm{s}$. For GNRs in dehydrated mucus ($>2$ wt. %), we observe a general trend that $D_T$ slows at smaller increments with increasing concentration and plateaus at $\sim 0.5\mu {\mathrm{m}}^2/\mathrm{s}$ [Fig. 3(a)]. In well-hydrated samples, $D_R$ retains a significant fraction of its value in solvent, above $\sim 3000{\mathrm{rad}}^2/\mathrm{s}$, then gradually decreases with further increases in mucus sample concentration. At the highest reported concentration, $D_R$ is calculated to be $\sim 1000$ and $\sim 1500{\mathrm{rad}}^2/\mathrm{s}$ in trials 1 and 2, respectively. Across all reported sample concentrations, outside of the solvent data, the variance in $D_R$ is less than half of that observed in $D_T$. $D_T$ intrasample variability is over 10% in all reported concentrations over 1 wt. %, whereas $D_R$ variability remains under 10%.

Fig. 3

GNR diffusion in hBE mucus sample. In the top panels of (a), $D_T$ and $D_R$ are plotted as a function of mucus solids concentration. In many cases, error bars are not visible due to small standard deviation size. Zero concentration represents GNR diffusion in solvent (DPBS). In the bottom panel, intrasample variability is presented as the percent ratio of the standard deviation to the mean for both $D_T$ and $D_R$. (b), (c) Co-polarized (HH) and cross-polarized (HV) B-mode images of the 0.5 and 6.4 wt. % hBE samples are displayed along with corresponding $D_T$ and $D_R$ values within each ROI derived from B + M-mode images; excluded ROIs based on thresholding criteria described in the methods are displayed as the background color. (c) Multiple ROIs are rejected due to air bubbles (indicated by blue arrows) in the 6.4 wt. % sample.

Overall, the trends in $D_T$ and $D_R$ versus mucus concentration closely mirror those observed in PEO solutions. If we define, as for the PEO experiments above, a weakly constrained regime of GNR diffusion as that when $D_T$ is $\ge 1/10$ the value in solvent, GNRs transition to being strongly constrained when mucus concentration exceeds $\sim 4.5$ wt. %. However, $D_R$ continues to be sensitive to mucus concentration in this strongly constrained regime. Also, similar to results in PEO, diffusion colormaps of the highest concentration sample in Fig. 3(c) show that there are multiple ROIs where $D_R$ is defined while $D_T$ is not (criterion 4). We note that samples in trials 1 and 2 were mixed under slightly different temperatures, potentially contributing to the observed differences in $D_T$ and $D_R$. The higher concentration samples are prone to air bubbles during the mixing process, an example of which can be seen in Fig. 3(c). Large areas of rejected ROIs are due to air bubbles.

Although $D_T$ measurements are a valid tool for assessing the nanostructure of well-hydrated mucus samples, these results with mucus point to $D_R$ being more sensitive to the concentration of dehydrated mucus samples. Incorporating our previously published translational diffusion in hBE mucus results in Ref. 27 and the above reported hBE experiments, we can define an assay for estimating the concentration of mucus given diffusion coefficient measurements. We first normalized all diffusion coefficients with respect to the diffusion coefficient in solvent measured on the same day, $D_{\text{norm}}=D/D_{\text{solvent}}$. A linear regression was then taken of $D_{T,\text{norm}}$ versus concentration over a well-hydrated range (0 to 2 wt. %), resulting in

To apply this assay, we collect $D_T$ and $D_R$ measurements in a mucus sample and solvent (saline), then compute weight percents from Eqs. (6) and (7). If the concentration falls between $-0.5\%$ and 2% for $D_T$, or between 1% and 8% for $D_R$ (extrapolating somewhat from the maximum measured value of 6.4 wt. %), we consider it valid and exclude values that fall outside these ranges. In the case where we have valid concentration values from both $D_T$ and $D_R$, we take the average. One limitation of this method is the potential variation in $R_H$ values among different batches of GNRs. To help account for potential nanorod size variability in $D_T$ measurements, we incorporated three GNR batches with distinct sizes when calculating the $D_T$ trendlines. The GNR batch used in this experiment has a size of $70\times 22\mathrm{nm}$ with a corresponding $R_H$ of 22 nm. The additional two batches of GNRs incorporated from our previous publication,²⁷ sized $83\times 22\mathrm{nm}$ and $62\times 10\mathrm{nm}$, had corresponding $R_H$ of 24 and 19 nm.

3.3.

Diffusion Coefficients of GNRs on In Vitro Mucus Under HTS Treatment

To mimic the airway epithelium physiology, Calu-3 cells were cultured under ALI conditions. The cell cultures were imaged with PS-OCT immediately before and over the course of 8.5 min after treatment with HTS. The mucus layer, which already contained GNRs introduced 6 h prior, was $\sim 300\mu \mathrm{m}$ thick before saline introduction. After topically introducing HTS with additional GNRs, the layer was $\sim 500\mu \mathrm{m}$ thick. We note that the bottom surface positions shift after the addition of the saline due to the refractive index of the liquid causing added optical delay in OCT. In Fig. 4(a), we observe $D_T$ of GNRs in the mucus layer of the ALI culture over depth and time. We define a depth of zero as the top of the epithelium where mucus is secreted from mucus goblets. Initially $D_T$ is rapid, with rates much higher than $7\mu {\mathrm{m}}^2/\mathrm{s}$ during HTS introduction. After 1 min, $D_T$ is measurable in ROIs close to the ALI but exhibits an increasing number of invalid ROIs in regions below, closer to the epithelium, likely due to high concentration that is outside the measurement range for $D_T$. Interestingly, at depths closer to the epithelium, $D_T$ starts to increase after 3 min, whereas it decreases near the top surface, resulting in a more homogenized value of $D_T$ over depth at 8.5 min.

Fig. 4

(a), (b) $D_T$ and $D_R$ in ALI cultures with HTS introduced are color mapped. (c) The diffusion coefficients are mapped to mucus concentration using trendlines produced from stationary mucus experiments. If an ROI had a valid measurement of both $D_T$ and $D_R$, the calculated concentrations were averaged for a final reported value.

Figure 4(b) reveals the corresponding $D_R$ of GNRs in the mucus layer of the ALI culture. Over the first minute, $D_R$ is largely undefined in between the ALI and the epithelium. The regions of interest that are quantified have high $D_R$, $>3500{\mathrm{rad}}^2/\mathrm{s}$. Beyond 2 min, $D_R$ in ROIs close to the surface remain consistently high. Underneath this area, a tall mass of high concentration mucus is revealed by $D_R$ measurements, in which the diffusion is hindered more than those at the ALI. Comparing Figs. 4(a) and 4(b), we see the complementary nature of $D_T$ and $D_R$, with $D_T$ having fewer invalid ROIs in regions of lower concentration and $D_R$ accessing higher concentration regions; at the same time, the overlapping ROIs of the two appear to follow the same depth- and time-resolved trends.

Using the trendlines defined in Eqs. (6) and (7) for relating $D_T$ and $D_R$ measurements to mucus concentration, in Fig. 4(c), we mapped the concentration of the mucus within the ALI cell culture in depth and time. By employing the new calibration method that exploits the complementary nature of the two diffusion coefficients, we produce a more spatially and temporally contiguous plot. Before HTS is introduced (at time periods $<\sim 20\mathrm{s}$), mucus appears hard-packed within a $400\mu \mathrm{m}$ layer. Immediately after HTS is introduced, we observe a $\sim 500\mu \mathrm{m}$ layer of liquid that appears to be mostly solvent, then a thin, highly concentrated layer of mucus near the ALI. Over time, high concentration mucus also appears lower down near the epithelium, and the depth-dependent concentration is generally heterogeneous. Between 4 and 8.5 min, the mucus appears to continue to mix and homogenize, likely as a result of ciliary activity on the epithelium, which we have found is more active in response to HTS than to isotonic saline.³⁵ Bearing in mind that “healthy” mucus is considered to be $\sim 2$ wt. %, we see that the mucus layer is nearing but has not quite reached, this value by the end of the experiment. In our previous publication,³⁵ we observed the reinitiation of the MCC in time-lapse OCT images during HTS treatment on hBEC. These results suggest the promising advantages of incorporating a combination of our advanced DS-OCT-based assay with regular OCT imaging for a comprehensive evaluation of the HTS treatment process in future applications.