Antibodies that bind to target receptors on the cell surface can confer conformation and clustering alterations. These dynamic changes have implications for characterizing drug development in target cells. This protocol utilizes confocal microscopy and image correlation spectroscopy through ImageJ/FIJI to quantify the extent of receptor clustering on the cell surface.
Confocal microscopy provides an accessible methodology to capture sub-cellular interactions critical for the characterization and further development of pre-clinical agents labeled with fluorescent probes. With recent advancements in antibody based cytotoxic drug delivery systems, understanding the alterations induced by these agents within the realm of receptor aggregation and internalization is of critical importance. This protocol leverages the well-established methodology of fluorescent immunocytochemistry and the open source FIJI distribution of ImageJ, with its inbuilt autocorrelation and image mathematical functions, to perform spatial image correlation spectroscopy (ICS). This protocol quantitates the fluorescent intensity of labeled receptors as a function of the beam area of the confocal microscope. This provides a quantitative measure of the state of target molecule aggregation on the cell surface. This methodology is focused on the characterization of static cells with potential to expand into temporal investigations of receptor aggregation. This protocol presents an accessible methodology to provide quantification of clustering events occurring at the cell surface, utilizing well established techniques and non-specialized imaging apparatus.
The development of therapeutic antibodies has shown remarkable success in the treatment of multiple tumor types1. The recent advancements of antibody-drug conjugates (ADC) as delivery mechanisms for cytotoxic compounds has expanded the requirements for understanding the dynamics of antibody:receptor interactions at the cell surface2. Following the successful targeting of an antibody to the cell surface receptor, these complexes can induce similar aggregation patterns to those observed in ligand:antibody interactions3. Alterations in receptor aggregation can induce changes to the membrane and result in the internalization of the receptor and its removal from the cell surface. In the context of an antibody-drug conjugate, this process subsequently releases the cytotoxic payload into internalized endosomes and subsequently the cytoplasm, resulting in effective cell killing.
Confocal microscopy has provided an effective means of visualizing these important interactions of antibodies and their target receptors4. To explore the changes of aggregation of target molecule at the cell surface this protocol utilizes post processing of confocal microscopy images via a spatial image correlation spectroscopy (ICS) technique 5,6,7.
The foundation of image correlation spectroscopy is the observation that spatial fluorescence intensity fluctuations share a relationship to the density and aggregation state of the labeled structures. This relationship is established following the calculation of a spatial autocorrelation function of a captured image5.
All variants of image correlation spectroscopy require the calculation of an image autocorrelation. This is followed by fitting this function to a two-dimensional Gaussian curve for the extraction of quantitative aggregation state parameters contained within the image. In simple terms the calculation of an image autocorrelation involves comparing all the possible pixel pairs contained within an image and calculating the likelihood that both equally as bright as each other. This is visualized as a function of the distance and directions of pixel separation8.
The theoretical framework for the image correlation spectroscopy was established and defined by Petersen and Wiseman et al.5,6. In this protocol, the autocorrelation calculations are performed in Fiji/ImageJ as well as a spreadsheet application, the basis for the intensity fluctuation spatial autocorrelation function can be described as (Eq 1):
where F represents the Fourier transform; F−1 the inverse Fourier transform; F* its complex conjugate; and the spatial lag variables ε and η. In spatial ICS, as described in this protocol, the autocorrelation function can be calculated using a 2D fast Fourier transform algorithm7,9. The autocorrelation at zero spatial separation otherwise known as zero-lag, g11(0,0), provides the inverse mean number of particles present per beam area of the microscope. It can be obtained by fitting the spatial autocorrelation function to a two-dimensional Gaussian function (Eq 2):
As the pixels captured within an image are contained within a set area and these measurements do not extend to infinity, the term g∞ is used as an offset to account for long-range spatial correlations contained within the image. For molecular-sized aggregates, ω is the point-spread function of the microscope and described by the full-width at half-maximum of the spatial autocorrelation function. The area contained within the point spread function of the instrument can be calibrated through the use of sub-resolution fluorescent beads.
For the image correlation spectroscopy protocol described herein, the autocorrelation and mathematical functions required to complete ICS are performed using the open-source imaging-processing platform, Fiji10, a distribution of the ImageJ program11,12. Fiji/ImageJ utilizes the preinstalled fast Fourier transformation in the FFT Math function. This function reduces the compute time required of this calculation by reducing the range of data by a factor of two in each dimension13. As the 2D autocorrelation function is approximately symmetrical in x,y axis, a single line profile plot through the autocorrelation image can be used to measure the raw autocorrelation as a function of the spatial lag. Any zero-lag noise is removed prior to further calculation, with the resulting autocorrelation amplitude (peak value, g(0)T) corrected for background with the expression (Eq 3):
where Ib is the mean intensity from a background region excluding the cell. The cluster density, or density of fluorescent objects, is defined by (Eq 4):
In the protocol described herein, we further simply the calculation of cluster density (CD), with the assumption based on the observation that a normalized autocorrelation function will decay to a value approaching 1.0 with increasing spatial lag. With maximum spatial lag, there is no longer any correlation of fluorescence intensity values and thus without a correlation the calculations at this region are computing the value of an intensity multiplied by this intensity which is subsequently divided by the square of that intensity, which by definition is equal to 1.0. Thus, cluster density from a normalized autocorrelation function can be computed by subtracting 1.0 from the normalized autocorrelation function prior to taking its reciprocal (Eq 5):
Further calibration of the beam area can be performed to quantitate the number of clusters contained within the area of the point spread function of the instrument. This calibration must be performed using the same optical conditions used during the image correlation spectroscopy analysis.
1. Seeding of Adherent Cell Lines for Imaging Experiment
2. Primary Antibody Incubation
3. Secondary Antibody Incubation
4. Confocal Microscopy of Samples
NOTE: The settings used for confocal analysis will differ depending on the equipment and fluorophores used in the individual experiment. Modern confocal instruments provide mechanisms for saving all imaging parameters used in the experiment in the native image file format of the instrument. It is important that these settings are recorded to provide for appropriate replication of experimental conditions across different periods of data acquisition.
5. Calculation of Image Correlation Spectroscopy
NOTE: The open-source imaging-processing platform, Fiji, a distribution of the ImageJ program, is required to perform the autocorrelation and mathematical functions essential to ICS. This distribution of ImageJ is recommended as it includes numerous pre-installed plug-ins and an easier update architecture that can perform operations across multiple imaging experiments.
6. Batch Processing of ICS datasets
NOTE: The establishment of a FIJI/ImageJ macro is recommended to replicate the procedures described in the above protocol. This allows for the consistent and rapid analysis of multiple image files during image correlation spectroscopy analysis.
7. Protocol Extensions: Calculation of Beam Area of Microscope
NOTE: The optical transfer function of the microscope ensures that even molecular-sized objects appear as images with a radius of about 200-300 nm in the x-y plane. The ICS protocol allows determination of the point-spread function by performing steps 4-5 on sub-resolution fluorescent beads. Specifically:
8. Protocol Extensions: Calculation of Clusters Per Unit of Area
A successful image correlation spectroscopy experiment is dependent on the proper application of treatment controls. These treatment groups include the examination of the fluorescence in no primary antibody and no secondary antibody treatment groups. Once the optimum confocal laser settings are established for an experiment, images must be captured of these control groups to confirm the lack of non-specific fluorescence within the sample. For many adherent cancer cell lines, the identification of regions suitable for ICS is limited by the lack of flat surfaces. This is compensated by the ability to capture a large field of view of multiple cancer cells in a single confocal image. This provides the ability to generate the required sampling to perform statistical comparisons between treatment groups.
In this representative experiment, A431 EGFR positive epidermoid carcinoma cells, were stimulated with 100ng/mL of EGF ligand for 10 min at room temperature to induce EGFR clustering. ICS analysis was performed on both EGF stimulated (Figure 1A) and unstimulated (Figure 1D) cells following incubation with the EGFR targeting humanized monoclonal antibody, cetuximab. Subsequently, a 64×64 pixel crop of a region (yellow box) contained within the cell membrane (Figure 1B and 1E) was processed utilizing the autocorrelation function was contained in the open-source image-processing platform, FIJI. This resulting autocorrelation image was normalized by subtracting the background fluorescence prior to dividing by both the square of average intensity of the normalized region of interest as well as the square of the number of pixels in the cropped region (Figure 1C and 1F). The line function tool was used to plot the profile of the point-spread function of this image (Figure 1G and 1H). The peak value of these profiles were transposed to a spreadsheet application for further data processing. A FIJI macro was generated to replicate these FIJI processing steps for rapid processing of a series of cells (n=6) from both treatment groups following cetuximab immunocytochemistry (Figure 1I). In this experiment, the EGFR receptor was identified in 5.84 ± 0.85 clusters per beam area on the surface of the A431 stimulated cells versus 14.94 ± 1.62 clusters per beam area in the unstimulated cells populations. With the assumption that the experimental conditions used for this analysis limit rapid receptor turnover, the EGFR cluster density in stimulated A431 cells decreased 2.56-fold. When conditions are compared for the aggregation state of the same target molecule in conditions when target molecule turn-over is limited, a lower number of clusters per beam area indicates increased target aggregation. Following EGF ligand stimulation EGFR aggregation increases 2.56-fold as expected in this system. The generation of a FIJI macro using the steps described in this protocol allows for the statistical comparison of varied treatments or environmental differences and their effects on receptor aggregation.
Figure 1: Image correlation spectroscopy to detect fluorescently labeled clusters on the cell surface of adherent cancer cells. Confocal microscopy image of representative EGF stimulated (A) or unstimulated (D) adherent A431 epidermoid carcinoma cells labeled with cetuximab primary antibody targeting cell surface EGFR receptor with Alex Fluor 488 secondary antibody. Stimulated cells were treated with 100 ng/mL EGF for 10 min at room temperature to induce EGFR clustering prior to fixation and immunocytochemistry. Cell membrane containing crop regions (yellow box) with pixel dimensions (64 x 64 [2n]) were used to perform ICS analysis (B and E). Following normalization of the autocorrelation function in FIJI/ImageJ, the line tool (yellow) was used to measure the point spread function (C and F) The profile of this line function is plotted to detect the peak value (G and H). EGF stimulation was observed to induce a 2.56-fold decrease in detected EGFR clusters per beam area 5.84 ±0.85 compared to 14.9 5±1.62 to unstimulated cells (n=6) (I). With the assumption, the experimental conditions used in this analysis maintained a constant number of cell membrane receptors, this represents an increase in EGFR aggregation following EGF stimulation. Scale Bars = 10 µm (A and D); 1 µm (B-C and E-F). Box Plot displayed using Tukey Style with whiskers representing maximum and minimum observed values no more than 1.5 × interquartile range (IQR). Clusters per beam area ± Standard Error of the Mean (SEM). Please click here to view a larger version of this figure.
The technique of image correlation spectroscopy (ICS) that we describe in this protocol uses standard confocal microscopes without the need for specialized detectors. The ICS technique described utilizes well-established immunocytochemistry methods to provide for rapid sampling of multiple treatment conditions for increased statistical analysis. This methodology does so with a slight reduction in absolute precision as compared to alternative single molecule techniques based on correlation of fluorescence fluctuations of mobile molecules diffusing through a confocal volume, such as fluorescence correlation spectroscopy (FCS)15. The more specialized FCS approach also requires high sensitivity photon counting detectors such as avalanche photodiode (APD) or GaAsP detectors together with dedicated specialized software that is not routinely available on standard confocal instruments. A comprehensive analysis of the precision of ICS determined this technique can accurately measure the number of clusters where >1 particle is present within the beam focal spot of the laser scanning microscope7.
The collection of appropriate fluorescent images is of critical importance for a successful ICS analysis. The region of interest must contain a high signal to background fluorescent ratio with laser and gain intensities adjusted to remove any signal saturation. The cellular region captured must be oversampled with a pixel size of approximately 50 nm. This region must be homogeneous, with careful attention payed to avoid cellular edges or junctions which can increase artifacts in the autocorrelation calculation. One of the major pitfalls of an ICS approach is the sensitivity to heterogeneous cellular surfaces. Thus, limiting the region captured to the maximum flat surface, regardless of cell shape will improve the accuracy of any ICS analysis.
Multiple imaging modalities have been explored to study the aggregation of receptor complexes. Electron microscopy (EM) provides high-resolution spatial assessment of protein clustering but operates under extremely high vacuum and is not applicable to studies requiring temporal resolution. EM is further limited as its harsh sample preparation is prohibitive for living tissue investigations5. To overcome the inherent resolution limitations of standard confocal microscopy, Förster resonance energy transfer (FRET) can be used to calculate the spatial organization of structures labeled with overlapping donor and acceptor fluorophores pairs16. Despite its sensitivity, FRET does not however provide information regarding the size of the aggregates under investigation6. The development of various super resolution fluorescence microscopy methods provide exciting opportunities to resolve cellular objects not possible with standard resolution confocal methods17,18. The expense and expertise required of such technologies and instruments, limits their current commonplace availability.
With increases in sensitivity and low photo-toxicity in confocal instrumentation, image correlation spectroscopy can be expanded to explore temporal changes of receptor aggregation in living cells in time-lapse experiments. This can be juxtaposed with increasing the laser power used to deliberately induce photo bleaching of individual samples with subsequent ICS assessment (pbICS) to tease apart the higher-order aggregation present in some receptor complexes9. A further adaption, image cross-correlation spectroscopy (ICCS) provides an analysis of the co-localization of two membrane based proteins through the analysis of image pairs for each fluorescently labeled protein19. The strength of this image correlation spectroscopy methodology is that it leverages techniques well established in the field and provides a freely available analysis pipeline to quantitative the highly dynamic events occurring at the cell membrane.
The authors have nothing to disclose.
The authors acknowledge funding support from NHMRC (Fellowship 1084178 and Grants 1087850, 1030469, 1075898 (AMS)), Cancer Australia, Ludwig Cancer Research, John T Reid Trusts, Cure Brain Cancer Foundation, La Trobe University and the Victoria Cancer Agency. Funding from the Operational Infrastructure Support Program provided by the Victorian Government, Australia is also acknowledged.
Nunc Lab-Tek II Chambered Coverglass – 8 well | ThermoFisher Scientific | 155409 | |
DPBS, no calcium, no magnesium | ThermoFisher Scientific | 14190144 | |
Fetal Bovine Serum | ThermoFisher Scientific | 10099141 | |
TrypLE Express Enzyme (1x), no phenol red | ThermoFisher Scientific | 12604021 | |
Penicillin-Streptomycin (10,000 U/mL) | ThermoFisher Scientific | 15140122 | |
GlutaMAX Supplement | ThermoFisher Scientific | 35050061 | |
Goat anti-Human IgG (H+L) Cross-Adsorbed Secondary Antibody, Alexa Fluor 488 | ThermoFisher Scientific | A11013 | |
TetraSpeck Fluorescent Microsphere Standards 0.1µm | ThermoFisher Scientific | T7279 | |
Cetuximab | Merck Serono | 3023715501 | |
Parafilm M 38mx100mm | Merck Millipore | BRND701605 | |
16% Paraformaldehyde (formaldehyde) aqueous solution | ProSciTech | C004 | |
Recombinant Human EGF | R&D System | 236-EG |