|

CAD-LB: Colocalization Detection Guide

BySam Walker

Introduction

In this post I will review work I have completed to further characterize the CAD-LB platform. In this set of experiments combinations of nanoparticle species were used to define colocalization limitations. More specifically, two nanoparticle species each labeled with a unique fluorescent dye (red or blue) were used to simulate pan EV and antibody signals. As with quantification, these limitations are dependent on NPN pore density and the optical resolution of our imaging system. In EV labeling experiments, high densities of fluorescent pan EV and antibody signals will inevitably result in nearby capture and false colocalization detection. As a result of this work the expected rates of false positive, as well as false negative (using multicolor nanoparticles), colocalization detection inherent to CAD-LB are defined.

Experimental Approach

Red and blue nanoparticle species were mixed in a defined manner to elucidate rates of false positive detection for fully independent markers. The red nanoparticle input was held constant (105) simulating a fluorescent pan EV signal, while the blue nanoparticle input spanned several orders of magnitude (102 – 105) simulating a variable fluorescent antibody signal. TetraSpeckTM microspheres were prepared at a single input (105) and used to simulate a true positive colocalization event. Nanoparticle solutions were processed using CAD-LB as described previously. Colocalization events identified in red/blue nanoparticle mixture experiments were classified as false positives, while colocalization events missed in TetraSpeckTM microsphere experiments were classified as false negatives.

Results & Discussion

Representative regions of interest (ROIs) from nanoparticle experiments are presented in Figure 1A-I. Panels A-F depict red/blue nanoparticle mixture experiments and panels G-I depict TetraSpeckTM microsphere experiments. In panels A-C, the variable blue nanoparticle input (102 – 104) and constant red nanoparticle input (105) can be visualized. In panels D-F, smaller ROIs from the blue, red, and merged channels are displayed for experiments using 104 blue and 105 red nanoparticles. Instances of false positive colocalization detection can be visualized in panel F and are indicated with arrows. In panels G-I, smaller ROIs from green, red, and merged channels are presented for experiments using 105 TetraSpeckTM microspheres. An instance of false negative colocalization detection can be visualized in panel I and is indicated with arrows in panels G-I.

Figure 1. Nanoparticle Colocalization Control Experiment Images. (A-F) False positive colocalization control experiments were completed using two nanoparticle species labeled with unique fluorescent dyes (red or blue). The red nanoparticle input was held constant (10^5) and the blue nanoparticle input spanned several orders of magnitude (10^2 – 10^5). (A-C) Representative regions of interest (ROI = 119 µm x 119 µm) in confocal microscope images taken following nanoparticle capture on NPN membranes. (A), (B), and (C) correspond with nanoparticle inputs of 10^5 red and 10^2 blue, 10^5 red and 10^3 blue, and 10^5 red and 10^4 blue, respectively. Scale bars = 10 µm. (D-F) Representative region of interest (ROI = 39 µm x 39 µm) in confocal microscope images taken following capture of 10^5 red and 10^4 blue nanoparticles. (D), (E), and (F) correspond with blue channel, red channel, and blue and red channel overlaid images, respectively. Arrows in (F) indicate instances of false positive colocalization detection. Scale bars = 2 µm. (G-I) Representative region of interest (ROI = 39 µm x 39 µm) in confocal microscope images taken following capture of 10^5 TetraSpeckTM microspheres. (G), (H), and (I) correspond with green channel, red channel, and green and red channel overlaid images, respectively. Arrows in (G-I) indicate an instance of false negative colocalization. Scale bars = 2 µm. All captured nanoparticles were imaged with a fluorescence microscope.

The false positive and false negative colocalization detection rates were quantified and are presented in Figure 2. Relative to the 105 red nanoparticle input, the rate of false positive colocalization scaled accordingly with the blue nanoparticle input (105 : 4.759%, 104 : 0.482%, 103 : 0.050%, 102 : 006%) (Figure 2A). This finding is expected and enumerates the colocalization specificity of CAD-LB. Relative to the variable blue nanoparticle input, the rate of false positive colocalization was constant (~4%); this finding is also expected since the red nanoparticle input was held constant. The false negative colocalization rate returned from TetraSpeckTM experiments was ~1.7% (Figure 2B), corresponding to favorable colocalization sensitivity of 98.3%.

Figure 2. Nanoparticle Colocalization Control Experiment Quantification. (A) Measured colocalization rates are presented relative to both species’ inputs (indicated by color). (B) False positive colocalization rates relative to red nanoparticle input presented in terms of blue : red nanoparticle ratio (%). Larger data points and tick marks shown on y-axis correspond with false positive colocalization curve completed with 10^5 red nanoparticle background; smaller data points correspond with false positive colocalization curve completed with 10^4 red nanoparticle background. Empty data point on 10^4 red background curve represents a predicted value using data associated with 100% and 10% blue : red nanoparticle ratio inputs. A single input (10^5) of TetraSpeckTM fluorescent microspheres labeled with multiple fluorescent dyes was used to determine the true positive/false negative colocalization detection rate (presented with horizontal line). Quantification and colocalization analyses were completed using an open-source ImageJ plugin (ComDet v.0.5.5). All data points represent mean ± SEM (n = 3).

Next, the rates of false positive colocalization detection for a constant 104 red nanoparticle input were determined. For the same blue : red nanoparticle ratios (100%, 10% and 1%), the 104 red nanoparticle input reduced false positive detection rates by roughly an order of magnitude (Figure 2B). This, again, is expected since lowering the total nanoparticle density decreases the likelihood of false positive colocalization. It should be noted that for experiments using 102 blue and 104 red nanoparticles no instances of colocalization were observed. Using false positive rates associated with blue : red nanoparticle ratios of 100% and 10% to predict the behavior of 1%, we anticipate a false positive rate of ~0.005% which corresponds with one false positive colocalization event occurring across 54 field of views (FOVs). In these experiments 15 FOVs (5 FOVs in 3 devices) were analyzed, and thus, our finding of 0% is consistent with the predicted false positive rate.

As a result of these experiments colocalization guides were generated for EV inputs of 105 (Figure 3A) and 104 (Figure 3B).

Figure 3. Nanoparticle Colocalization Control Experiment Guides. (A) Guide for colocalization detection using blue (10^2 – 10^5) and red (10^5) nanoparticles to simulate antibody and pan EV signals, respectively. Green and red regions depict putative regions of true and false positive colocalization detection, respectively. (B) Guide for colocalization detection using blue (10^2 – 10^4) and red (10^4) nanoparticles to simulate antibody and pan EV signals, respectively. Green and red regions depict putative regions of true and false positive colocalization detection, respectively. Empty data point on curve represents a predicted value using data associated with 10^3 and 10^4 blue nanoparticle inputs. All data points represent mean ± SEM (n = 3).

Putative regions of true positive (green) and false positive (red) colocalization detection have been denoted to explicitly define CAD-LB’s capacity for biomarker interrogation. Of course, in EV experiments isotype controls will be used to determine levels of nonspecific binding which will be used as correction factors. False positive colocalization rates derived from the observed signal ratio (antibody : pan EV) provide an independent specificity metric that is based on antibody signal prevalence. For example, when 104 antibody and 105 pan EV counts are observed, the false positive detection threshold is ~0.48%. This will be particularly useful since antibody signal density is variable; reagents are uniquely produced which may result in different nonspecific binding and aggregation behavior.

Similar Posts