We describe the use of quantitative immunoblotting to validate immunofluorescence histology coupled with image analysis as a means of quantifying a protein of interest in formalin-fixed, paraffin-embedded (FFPE) tissue samples. Our results demonstrate the utility of immunofluorescence histology for ascertaining the relative quantity of biomarker proteins in routine biopsy samples.
Quantification of proteins of interest in formalin-fixed, paraffin-embedded (FFPE) tissue samples is important in clinical and research applications. An optimal method of quantification is accurate, has a broad linear dynamic range and maintains the structural integrity of the sample to allow for identification of individual cell types. Current methods such as immunohistochemistry (IHC), mass spectrometry, and immunoblotting each fail to meet these stipulations due to their categorical nature or need to homogenize the sample. As an alternative method, we propose the use of immunofluorescence (IF) and image analysis to determine the relative abundance of a protein of interest in FFPE tissues. Herein we demonstrate that this method is easily optimized, yields a wide dynamic range, and is linearly quantifiable as compared to the gold standard of quantitative immunoblotting. Furthermore, this method permits the maintenance of the structural integrity of the sample and allows for the distinction of various cell types, which may be crucial in diagnostic applications. Overall, this is a robust method for the relative quantification of proteins in FFPE samples and can be easily adapted to suit clinical or research needs.
The need to quantify proteins in formalin-fixed, paraffin-embedded (FFPE) tissue biopsy samples exists in many clinical fields. For example, quantification of biomarker proteins in routine biopsy specimens is used to elucidate prognosis and inform treatment for cancer patients1. However, current methods are typically subjective and lack validation.
Immunohistochemistry (IHC) is used routinely in pathology laboratories and generally depends on a primary antibody directed at the target protein and a secondary antibody conjugated with an enzymatic label such as horseradish peroxidase2. Conventional IHC is sensitive, can make use of minute samples and preserves the morphological integrity of tissue samples thereby permitting assessment of protein expression within its relevant histological context. However, because the chromogenic signal generated by IHC is subtractive, it suffers from a relatively narrow dynamic range and offers limited potential for multiplexing2,3,4. Matrix-assisted laser desorption/ionization mass spectrometry imaging (MALDI-MSI) preserves morphological integrity. However, this developing technology is associated with modest morphological resolution and requires significant calibration and normalization, impairing its feasibility for routine clinical use5,6,7. Alternative techniques to quantify protein in tissue samples include immunoblotting8, mass spectrometry9,10,11, and enzyme-linked immunosorbent assay (ELISA)12, each of which begins with a homogenized lysate of sample tissue. Primary tissue samples are heterogeneous in that they contain a multitude of cell types. Therefore, techniques that entail homogenizing the samples do not permit quantification of a protein in a particular cell population of interest such as cancer cells.
Like IHC, IF is applicable to small FFPE samples and permits the retention of histological integrity13. However, thanks to the additive nature of fluorescence signals, IF is amenable to the application of multiple primary antibodies and fluorescent labels. Thus, a protein of interest may be relatively quantified within specific cells or cellular compartments (for example, nucleus versus cytoplasm) defined using other antibodies. Fluorescence signals also have the advantage of a greater dynamic range13,14. The superiority, reproducibility, and multiplexing potential of IF applied to FFPE samples has been demonstrated13,14,15.
Herein we describe the use of quantitative immunoblotting using established cell lines as a gold standard to ascertain the quantitative nature of IF coupled with computer-assisted image analysis in determining the relative abundance of a protein of interest in histological sections from FFPE tissue samples. We have applied this method successfully in a multiplex approach to quantify biomarker proteins in clinical biopsy samples16,17,18.
Approval to use primary human tissue samples was obtained from the Health Sciences and Affiliated Teaching Hospitals Research Ethics Board (HSREB) at Queen's University.
1. Building a Cell-line Tissue Microarray (TMA)
2. Sample staining by Immunofluorescence
3. Quantitative Immunoblotting of Cell Lines
This protocol was used to confirm the ability of IF to determine the relative quantity of the anti-apoptotic protein Bcl-2 in cell lines made into FFPE tissue blocks. Quantifying Bcl-2 selectively in cancer cells can elucidate oncogenic mechanisms and can be useful in pathological diagnosis and in informing clinical management decisions24. More specifically, Bcl-2 plays a role in proper B-lymphocyte development and its expression is commonly investigated in the context of lymphoma25,26,27. Figure 1 outlines the steps involved in the protocol. In an initial IF optimization step, various dilutions of the primary anti-Bcl-2 antibody were tested on human tonsil tissue using an automated immunohistology stainer as described in step 2.1. Figure 2 contains images of the stained and scanned histology slides of human tonsil tissue that each received a different dilution of antibody. It can be seen that 1:50 is the optimal dilution that yielded strong signal and little background fluorescence. This dilution was then used on the cell line TMA as described in step 2.2. The TMA was also stained using DAPI to identify nuclei. A tyramide-based signal amplification kit was used to label Bcl-2 with a Cy5 fluorophore. Image analysis was used to quantify the cytoplasmic Cy5 fluorescence signal attributed to Bcl-2 in each cell line. Representative images of the staining can be seen in Figure 3. The immortalized cell lines chosen for this experiment included a variety of lymphoid-derived cell lines, namely 697, JeKo-1, Jurkat, RCH-ACV, Granta-519, REH, and Raji, in addition to HeLa, derived from a cervical carcinoma. The HeLa cells are known to express Bcl-2 at a very low level28.
Based on an initial immunoblot of all eight cell lines, Granta-519 was determined to have the greatest abundance of Bcl-2 (not shown). Serial dilutions of the Granta-519 lysate were used in a subsequent immunoblot to find the linear dynamic range of the Bcl-2 and GAPDH (loading control) signals (Figure 4A). This immunoblot was exposed using the digital scanner for varying lengths of time. Densitometry using image analysis software was used to quantify the signal from each band, and these values were plotted against the amount of protein loaded (Figure 4B). From the data in Figure 4B-top, the dynamic range for Bcl-2 in this assay spans from a band intensity of nearly zero to 7500 (arbitrary units, blue line). The two higher exposure times fit a quadratic and non-linear equation, suggesting overexposure and saturation of the signal intensity. The range for GAPDH is from 3000 to 6500 (arbitrary units, Figure 4B-bottom). Values below 3000 (arbitrary units) dropped precipitously even when using a relatively low exposure time. A long exposure clearly results in saturation. From these graphs, it was determined that 12 µg would be a reasonable amount of protein to load when performing the immunoblot with all cell lines, since this amount of protein yielded Bcl-2 intensity values within the linear range for the Granta-519 cells, reducing the risk of overexposure for all other cell lines.
A second immunoblot of all cell lines was then performed as in step 3.4 and can be seen in Figure 5A. This blot was required since the initial blot contained bands with a signal intensity outside of the linear range. Image analysis software was used to determine the signal intensity of each band in the new blot. Only intensity values that were within the dynamic ranges determined above were used. The arrows on the right of Figure 4B demonstrate representative intensity values that were used and show where they fit within the linear range. The ratio of Bcl-2:GAPDH was then calculated for each cell line. This ratio, along with the fluorescence signal from IF can be seen in Figure 5B. A Pearson correlation test demonstrated that the intensity ratios from immunoblotting were strongly and positively correlated with the intensity readings from quantitative IF (r = 0.983, p < 0.001; see Figure 5C).
Quantitatively assessing the amount of Bcl-2 proved to be particularly difficult as there was a wide range of expression of this protein across the eight tested cell lines. Using a long enough exposure to capture the signal of the low Bcl-2-expressing cell lines (>1 s) made it difficult to remain in the dynamic range for the high Bcl-2-expressing cell lines. In an attempt to quantify the faint bands produced by cell lines such as HeLa and Raji, several different exposure times, from 0.1 s to 5 s, were captured to determine the longest exposure time that could be used while remaining in the dynamic range for cells such as Granta-519 (see Figure 6). The nature of this immunoblotting technique limits the accuracy of signal detection as one approaches noise, suggesting it is optimally used to quantify proteins found at intermediate to high expression levels.
Figure 1: Protocol workflow diagram. Immunofluorescence (IF) on a cell line tissue microarray (TMA) was run in parallel to quantitative immunoblotting (IB) of the same cell lines. Signals from each cell line are compared by Pearson correlation to validate the quantitative ability of the IF protocol. Please click here to view a larger version of this figure.
Figure 2: Testing and optimizing immunofluorescence (IF) protocol. Sections of tonsil were incubated with different dilutions of anti-Bcl-2 antibody (1:50, 1:100, and no primary antibody). The slides were incubated with HRP-conjugated secondary antibody, the signal was amplified using a tyramide-Cy5 conjugate, and the slides were stained with DAPI. The slides were scanned at 20X magnification using filter sets appropriate to the maximal excitation/emission peaks of 350/470 nm and 649/666 nm for DAPI and Cy5, respectively. Respective exposure times were 160 ms and 200 ms. The field of view is centered over the same lymphoid follicle, which is expected to be Bcl-2 negative in the (central) germinal center and positive in the (peripheral) mantle zone. The 1:50 dilution gave a stronger specific fluorescence signal without increasing the nonspecific signal therefore it was used as the dilution going forward. Please click here to view a larger version of this figure.
Figure 3: Immunofluorescence (IF) using cell line tissue microarray (TMA) sections. Sections of the cell line TMA were taken and incubated with a 1:50 dilution of anti-Bcl-2 or with no primary antibody. The slides were incubated with secondary antibody, the signal was amplified using a tyramide-Cy5 conjugate, and the slides were stained with DAPI. The slides were scanned at 20X magnification using filter sets appropriate to the maximal excitation/emission peaks of 350/470 nm and 649/666 nm for DAPI and Cy5, respectively. Respective exposure times were 250 ms and 625 ms. These representative images indicate that staining was successful, and that there is a range of Bcl-2 expression across the cell lines. The "no Bcl-2 antibody" negative control slide did not demonstrate any Cy5 signal, as expected. The representative image is from the 697 cell line. A core of hyperplastic tonsil tissue included on the TMA demonstrates an equivalent pattern to that shown in Figure 2. The mean fluorescence intensity (MFI) for each cell line was determined by quantifying the Cy5 fluorescence signal using image analysis software. Please click here to view a larger version of this figure.
Figure 4: Determining the dynamic linear range of the immunoblot (IB) signals. (A) IBs of serial dilutions of Granta-519 lysate. Each blot was exposed for a variety of times to find the linear dynamic range. (B) Band intensity for Bcl-2 (top) and GAPDH (bottom) versus amount of total protein loaded for IB in (A). Band intensities were quantified using the image analysis software ImageJ. For Bcl-2 (top), the signals remained linear throughout the range of band intensities when a low exposure (exp.) time (0.2 s) was used (intensities from 0-7,200) as supported by a Pearson's correlation coefficient (r value) of 0.994 (p < 0.001), but not for greater exposure times of 1 s and 2 min. For GAPDH (bottom), the data remained linear above an intensity of 3000 (arbitrary units) for low (0.2 s) and medium (2.2 s) exposure times as supported by Pearson's r values of 0.994 (p < 0.001) and 0.992 (p < 0.001), respectively, but not for a high exposure time (7 s). Arrows on the right indicate the band intensity for the specific cell line identified from the immunoblot in Figure 5A. Please click here to view a larger version of this figure.
Figure 5: Quantitative immunoblotting (IB) of immortalized cell lines and comparison to immunofluorescence (IF). (A) IB of all cell lines. 12 µg of total protein was loaded into each well. (B) IB and IF signal intensities for each cell line. The IB signal is the ratio of Bcl-2:GAPDH band intensity from the blot in (A) as quantified by ImageJ. The IF signal is the fluorescence intensity of each cell line core on the TMA (Figure 3) as determined by image analysis. (C) Scatterplot of IF signal versus IB signal. Each data point represents a given cell line. The two methods produced linearly correlated results with a Pearson r value of 0.984 (p < 0.001). Please click here to view a larger version of this figure.
Figure 6: Finding the optimal exposure time for quantitative immunoblotting (IB). IB of all cell lines is shown. Twelve micrograms of protein were loaded in each well. For each IB, various exposure times were used to capture an image where all bands remained within the linear dynamic range. The exposure times of the Bcl-2 and GAPDH strips respectively were: (A) 0.1 s and 0.1 s, (B) 1 s and 0.2 s, and (C) 5 s and 5 s. Panels A and C show examples of exposures where at least some of the bands on the blot were too faint or too strong, respectively. Please click here to view a larger version of this figure.
We have described a method that makes use of quantitative immunoblotting (IB) to demonstrate the utility of immunofluorescence (IF) for ascertaining the relative abundance of a target protein in FFPE tissue samples. Current protein quantification methods are limited by their categorical nature, such as chromogenic IHC2,3, or by the need to homogenize samples, preventing investigation into the sample structure and cell populations, such as with IB and mass spectrometry8,9,10,11. Quantitative IF can transcend these limitations if the method is validated and applied carefully. By comparing the IF readouts to quantitative iIF of immortalized cell lines, we were able to validate the semi-quantitative nature of the IF approach.
Primary tissue samples from patients or experimental animals are heterogeneous in that they contain multiple cell types. The application of several primary antibodies in multiplex IF permits the quantification of one or more proteins of interest in specific cell types or cell compartments4,13. The optimization and application of multiplex IF is beyond the scope of this article but is outlined in the following papers16,17,18,29. The broad philosophy is to label cell populations and only quantify the protein of interest in the desired cell compartment, for example nuclei or cytoplasm, in the desired cell type. Once the quantitative nature of IF with a particular antibody has been verified by quantitative IB, the protocol can be upscaled to allow for rapid, high-throughput protein quantification in clinical samples. Clinical applications generally require the determination of relative, rather than absolute, protein abundance. However, the IF approach can be made formally quantitative if necessary. For example, a solution containing purified recombinant Bcl-2 protein could be prepared and quantified using a standard biochemical method. Serial dilutions could then be evaluated by quantitative IB and a standard curve generated to estimate the absolute protein content per cell in each cell line.
We applied our IF protocol to TMA sections representing biopsy samples from 66 cases of de novo diffuse large B-cell lymphoma. Our results demonstrated acceptable run-to-run reproducibility (Pearson r = 0.837) and the expected, strong association between "Bcl-2-positive" status determined subjectively by visual scoring of conventional IHC and elevated expression determined objectively by IF (p < 0.001). Furthermore, Bcl-2 abundance by IF correlated with Bcl-2 mRNA abundance in the same samples (Spearman ρ = 0.69, p < 0.001) and was significantly greater in cases with copy number gains of the BCL2 gene (p = 0.042) or translocation of BCL2 to the IGH locus (p = 0.004), as determined by fluorescence in situ hybridization. We obtained equivalent results in a separate cohort of cases. These findings provide additional evidence that supports IF as a valid method for determining the relative abundance of Bcl-2 in routine FFPE specimens. The use of IF to quantify several other biomarker proteins in primary samples has been described previously16,17.
Optimization of this protocol is a two-fold process. First, IF with primary antibodies used must be optimized (step 2.1). Commercial antibodies commonly suggest dilutions for IHC protocols which should be used as a starting point along with a dilution or two above and below this (i.e., if 1:100 recommended, add 1:50 and 1:150). After performing the IF and scanning the slide, determining which dilution is "optimal" entails visual inspection to identify the primary antibody dilution that generates the brightest signal without increasing background fluorescence. The second stage of optimization involves choosing the amount of protein to load for the immunoblot to ensure each band remains in the linear range (step 3.4). To ensure this, an immunoblot is performed using a serial dilution of the cell line that expresses the greatest abundance of target protein. From here, the band intensities versus protein amount can be plotted to determine the intensity ranges through which the signal remains linear. Since this range is established in the cell line with the most abundant target protein, for a given amount of protein all other cell lines will yield lower signals. When moving to an immunoblot of all cell lines (step 3.4), the dilution series is used to inform a protein amount to load that will yield strong signals without running the risk of overexposure beyond the linear range.
Despite the benefits of this protocol, the use of immunoblotting as a quantification verification method has its shortcomings. The primary concern revolves around the large range of expression for various proteins between specimens. It may be difficult to keep all samples within the limits of a linear range if some samples express the protein of interest to a much greater degree than others. In the representative results shown above, varying exposures were captured to get an image where the extremes (high and low) are both within the linear dynamic range (Figure 6). The greater than expected IF signals seen for the HeLa and Jurkat cells in Figure 5B may indicate either the lowest possible IF signal for this system, or that the IB signals have reached the bottom of the linear detection limit and are less accurate. Either way, this indicates that IF values at that range are less likely to be accurate in this system, and that the technique is optimal for medium- to highly-expressed proteins. Additionally, the use of HRP-conjugated secondary antibodies is not optimal for quantitative purposes, as discussed earlier with respect to IHC2,3. Using fluorescently tagged secondary antibodies for the immunoblot would provide a more robust confirmation that the IF is indeed linearly quantitative30,31, however, the Pearson r value of 0.984 obtained using the HRP-conjugated secondary antibodies in the above protocol was sufficient for the current purposes. Potential shortcomings associated with IF, such as autofluorescence and quenching vary based upon individual laboratory instrumentation, practices and tissue type, and can be minimized by various preventative measures32,33. A very small amount of autofluorescence in the representative results can be seen in no Bcl-2 antibody sample in Figure 3, however this amount is trivial in comparison to the remaining IF signals.
The need to accurately quantify proteins of interest from FFPE tissue samples pertains to clinical and research purposes across many biological fields. The protocol presented here outlines the use of quantitative IF on FFPE tissue sections and validates its semi-quantitative nature by immunoblotting of immortalized cell lines. This method allows quantification across a broad dynamic range, as well as the maintenance of the structural integrity of the tissue sample, which is not preserved in many other quantification methods3,8,11. Additionally, this protocol is easily adaptable and can be applied in research and clinical settings, especially for cancer-related research and prognosis.
The authors have nothing to disclose.
This work was partially funded by the Fredrick Banting and Charles Best Canada Graduate Scholarship (A.M.).
697 | DSMZ | ACC 42 | Cell line |
JeKo-1 | ATCC | CRL-3006 | Cell line |
Jurkat | ATCC | TIB-152 | Cell line |
RCH-ACV | DSMZ | ACC 548 | Cell line |
Granta-519 | DSMZ | ACC 342 | Cell line |
REH | DSMZ | ACC 22 | Cell line |
Raji | ATCC | CCL-86 | Cell line |
HeLa | ATCC | CCL-2 | Cell line |
Trypsin/EDTA solution | Invitrogen | R001100 | For detaching adhesive cells |
Fetal bovine serum (FBS) | Wisent Inc. | 81150 | To neutralize trypsin |
Neutral Buffered Formalin | Protocol | 245-684 | For fixing cell pellets |
UltraPure low melting point agarose | Invitrogen | 15517-022 | For casting cell pellets |
Mouse monoclonal anti-human Bcl-2 antibody, clone 124 | Dako (Agilent) | cat#: M088729-2, RRID: 2064429 | To detect Bcl-2 by immunoflourencence and immunoblot (lot#: 00095786 |
Ventana Discovery XT | Roche | – | For automation of immunofluorescence staining |
EnVision+ System- HRP labelled polymer (anti-mouse) | Dako (Agilent) | K4000 | For immunofluorescence signal amplification |
EnVision+ System- HRP labelled polymer (anti-rabbit) | Dako (Agilent) | K4002 | For immunofluorescence signal amplification |
Cyanine 5 tyramide reagent | Perkin Elmer | NEL745001KT | For immunofluorescence signal amplification |
Aperio ImageScope | Leica Biosystems | – | To view scanned slides |
HALO image analysis software | Indica Labs | – | For quantification of immunofluorescence |
Protease inhibitors (Halt protease inhibitor cocktail, 100X) | Thermo Scientific | 1862209 | To add to RIPA buffer |
Ethylenediaminetetraacetic acid (EDTA) | BioShop | EDT001 | For RIPA buffer |
NP-40 | BDH Limited | 56009 | For RIPA buffer |
Sodium deoxycholate | Sigma-Aldrich | D6750 | For RIPA buffer |
Glycerol | FisherBiotech | BP229 | For Laemlli buffer |
Bromophenol blue | BioShop | BRO777 | For Laemlli buffer |
Dithiothreitol (DTT) | Bio-Rad | 161-0611 | For Laemlli buffer |
Bovine serum albumin (BSA) | BioShop | ALB001 | For immunoblot washes |
Protein ladder (Precision Plus Protein Dual Color Standards) | Bio-Rad | 161-0374 | For running protein gel |
Filter paper (Extra thick blot paper) | Bio-Rad | 1703969 | For blotting transfer |
Nitrocellulose membrane | Bio-Rad | 162-0115 | For blotting transfer |
Trans-blot SD semi-dry transfer cell | Bio-Rad | 1703940 | For semi-dry transfer |
Skim milk powder (Nonfat dry milk) | Cell Signaling Technology | 9999S | For blocking buffer |
Tween 20 | BioShop | TWN510 | For wash buffer |
GAPDH rabbit monoclonal antibody | Epitomics | 2251-1 | Primary antibody of control protein (lot#: YE101901C) |
Goat anti-mouse IgG (HRP conjugated) antibody | abcam | cat#: ab6789, RRID: AB_955439 | Secondary antibody for immunoblot |
Goat anti-rabbit IgG (HRP conjugated) antibody | abcam | cat#: ab6721, RRID: AB_955447 | Secondary antibody for immunoblot (lot#: GR3192725-5) |
Clarity Western ECL substrates | Bio-Rad | 1705060 | For immunoblot signal detection |
Amersham Imager 600 | GE Healthcare Life Sciences | 29083461 | For immunoblot signal detection |
ImageJ software | Freeware, NIH | – | For densitometry analysis |