|

Characterization of the CAD-LB: Dynamic Range & Capture Efficiency

BySam Walker

Introduction

This post is meant to serve as an update for some of the more recent work completed on the CAD-LB (catch and display for liquid biopsy) platform. The CAD-LB is intended to be used as a digital biomarker assay for interrogating small extracellular vesicles (sEVs). Since transitioning our assay to the µSIM device we worked to rigorously characterize the performance and limitations of the CAD-LB. This post details some of this characterization as we define parameters like capture efficiency and dynamic range using input solutions of varying complexity.

Experimental Approach

Nanoparticle/sEV Capture

To remind everyone how we use the µSIM device, I have included the figure below.

Figure 1. Schematic of µSIM Nanoparticle/sEV Capture. (Top) Depiction of a fully assembled µSIM device. (Middle) Schematic of device featuring nanoporous membrane separated by well structure and microfluidic channel. (Bottom) Representation of experimental capture process.

In this configuration, one port of the µSIM device is blocked and a small volume (40 µL) of sample is injected into the other port. This way we create a dead-end-like filtration flow scheme in which the species of interest (nanoparticles/sEVs) is captured beneath the membrane.

Nanoparticle/sEV Imaging & Quantification

Fluorescent nanoparticles or fluorescently labeled sEVs are imaged with a fluorescence confocal microscope and the resulting fluorescent signals are quantified with an open-source ImageJ plugin. sEVs are labeled with carboxyfluorescein succinimidyl ester (CFSE) or MemGlow 488 which should label the lumen and lipid bilayer components, respectively.

Results & Discussion

In an effort to define the dynamic range of the CAD-LB, 100 nm fluorescent nanoparticles were captured, imaged and quantified. The quantification results and an accompanying error analysis are presented in Figure 2.

Figure 2 Nanoparticle Capture Experimental Results & Error Analysis. (a-c) Representative field of view (FOV = 206 µm x 206 µm) confocal microscope images taken following 100 nm nanoparticle capture. Counts are given for respective FOVs in bottom right corner of panels, scale bar = 5 µm. (d) Comparison of nanoparticles counted and nanoparticle input injected, total nanoparticle count is extrapolated from the average of 3 FOVs. The expected count is adjusted based on a systematic error inherent in nanoparticle capture experiments. (e) Counting error presented in terms of log10(expected count) – log10(captured count). The labels a-c denote nanoparticle inputs associated with images shown in panels (a-c). 100% capacity vertical lines indicate case in which the nanoparticle input is equal to the total number of membrane pores. (f) Comparison of counting error and theoretical interparticle spacing following capture. The dimensions of the optical resolution of the imaging system (R) and camera pixel size (P) are denoted with vertical lines. 1% and 100% capacity vertical lines indicate interparticle spacing at respective nanoparticle inputs. All error bars represent standard error of the mean (n = 3).

It appears that we are able to accurately quantify nanoparticle inputs over a range of ~102 – 106 (Figure 2d). Beyond inputs of 106 the CAD-LB incurs considerable error as the captured count falls short of the theoretical input. This can best be explained by visualizing the scenario depicted in Figure 2c; at higher inputs nanoparticle crowding occurs which decreases our ability to individually resolve nanoparticles. In terms of log10 error, the CAD-LB faithfully represents the order of magnitude of nanoparticle inputs over the range of ~102 – 106 (Figure 2e). Considering the theoretical average interparticle distance following capture, error greater than an order of magnitude is encountered when approaching dimensions associated with the optical resolution of the imaging system (Line R in Figure 2f) and the size of camera pixels (Line P in Figure 2f). A valuable finding is that the upper limit of the CAD-LB dynamic range is reached at ~1% membrane capacity, that is, the input count is 1% of the total pore count (~107).

It is important to note that there is a discrepancy between the theoretical input of nanoparticles and the captured count. The expected count was reduced after observing a systematic error spanning the linear region of the counting curve; the CAD-LB reported inputs that were consistently lower (~22.5%) relative to the theoretical input. To ensure that nanoparticles were not being lost post-capture we performed live imaging experiments, one of which can be viewed in the video below:

It is difficult to focus on the membrane during live capture due to the membrane bowing under pressure. At approximately second 12 of the video the membrane goes out of focus again, however, this movement is associated with the pressure being released as the pipette tip was ejected. After examining the frames immediately before and after the pressure was released, it is clear that virtually all captured nanoparticles are retained (Figure 3).

Figure 3 Pressure Release Effect on Nanoparticle Retention. Same representative field of view (FOV) confocal microscope image taken following 100 nm nanoparticle capture. (Left) FOV just prior to pressure release. (Right) FOV immediately following pressure release.

Quantification of these images confirms this as nearly identical counts are returned from both images. The abovementioned discrepancy could therefore be a result of the manufacturer’s over-estimation of nanoparticle concentration, loss associated with sample preparation or device dead volume, or some combination thereof. Following this, similar experiments were completed using purified sEVs collected from primary bladder epithelial (PBE) cell conditioned cell culture media. sEVs labeled with either CFSE or MemGlow were prepared at various inputs and captured and quantified as before (Figure 4).

Figure 4. CAD-LB Counting Curve of Purified sEVs. Solutions of purified sEVs (from primary bladder epithelial [PBE] cell conditioned cell culture media) were prepared at various inputs. PBE-derived sEVs were labeled with carboxyfluorescein succinimidyl ester (CFSE) or MemGlow 488, captured, imaged using a fluorescence microscope, and quantified using an open-source ImageJ plugin. Error bars represent standard error of the mean (n=3).
As a result of these experiments, several things are evident. Foremost, a similar dynamic range is achieved by the CAD-LB for purified sEVs and nanoparticles (~102 – 106). This once again means that we can comfortably work with inputs that correspond with ~1% membrane capacity. Second, it is important to qualify that this range is based on the quantification of fluorescent signals captured on the membrane. When comparing captured counts to theoretical inputs, which are based on an NTA measurement of the sEV stock solution, they are roughly an order of magnitude lower than expected. This disagreement may be due to several factors including the tendency for NTA to overestimate particle concentration and/or lower sEV labeling efficiency within the CAD-LB assay. Despite this, it is encouraging that both sEV general markers (CFSE & MemGlow) agree with each other over the observed range. The fact that these markers label different sEV components and result in such similar counting curves should instill confidence in CAD-LB sEV labeling and quantification.

Finally, counting curves were completed with sEVs collected from serum and plotted alongside the previously reported counting curves (Figure 5).

Figure 5. Summary of CAD-LB Dynamic Range. Solutions of nanoparticles, purified sEVs (from primary bladder epithelial [PBE] cell conditioned cell culture media), and serum-based sEVs were prepared at various inputs. The species of interest were labeled with carboxyfluorescein succinimidyl ester (CFSE), captured, imaged using a fluorescence microscope, and quantified using an open-source ImageJ plugin. Error bars represent standard error of the mean (n=3).
For the purpose of this summary figure the different solutions are compared based on the captured counts; that is, the curves were aligned along the x-axis such that captured count trends could be compared. We can now visualize the capability of the CAD-LB to process different particle sources. For instance, nanoparticle studies revealed that he CAD-LB incurs quantification error at inputs >106 due to the limitations of the imaging system. Furthermore, purified sEVs from conditioned media closely follow the trend of nanoparticles, however, serum sEVs produce a counting curve that plateaus in the ~104 range. This reduced dynamic range can likely be attributed to contaminating species in serum, like lipoproteins and protein aggregates, that compete for pore occupancy but are not fluorescently labeled.

In summary, we have taken steps to characterize the CAD-LB assay defining aspects including capture efficiency and dynamic range using various types of solutions. The next steps will include colocalization controls which will begin to assess the diagnostic capabilities of the CAD-LB, specifically probing its sensitivity and specificity.

Similar Posts