Film and CCD Imaging of Western Blots: Exposure Time, Signal Saturation, and Linear Dynamic Range

Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

Table of Contents Page 1. Introduction...... 1 2. Experimental Design ...... 2 3. Materials and Methods ...... 3 4. Results and Analysis ...... 3 4.1. Linear dynamic range, sensitivity, and limit of detection (LOD)...... 3 4.2. Experimental linear dynamic range is more limited than theoretical dynamic range...... 6 4.3. Impact of exposure time on linear dynamic range ...... 8 4.4. Signal-to-noise analysis...... 10 5. Discussion ...... 12 5.1 Choosing an appropriate detection method ...... 12 5.2 Impact of signal saturation and lower limit of detection...... 13 5.3 Enhancing Western blot reproducibility ...... 13 6. References...... 15

1. Introduction Linear dynamic range is a critical factor in quantitative analysis of Western blot data. In the cell, en- dogenous protein levels span a vast dynamic range (an estimated 4-10 orders of magnitude).1 Because protein levels vary so widely, accurate Western blot analysis requires detection methods that can accommodate a wide dynamic range and effectively capture the richness and complexity of the data. Many common detection methods cite a wide theoretical linear dynamic range. In practice, however, the available linear dynamic range may be much narrower and is generally limited by saturation of strong signals. Multiple exposures are usually captured for each blot, in an effort to mitigate the impact of signal saturation and determine the optimal exposure time with the widest linear dynamic range. But this practice can affect the reproducibility and reliability of data, because the chemiluminescent reaction is not constant and light output changes over time. It is important to identify and explore factors that influence linear dynamic range, to understand and minimize their impact on data analysis and interpretation.

Linear dynamic range is the span of signal intensities, from faintest to strongest, that displays a linear relationship between light emission from the sample and signal intensity recorded by the detector. This study investigates the linear dynamic range and limit of detection for several common Western blot Page 2 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

detection methods. Film exposure, a conventional charge-coupled device (CCD) imaging system (Imager B), and an alternative CCD system (Odyssey® Fc Imager) were tested and compared. Linear dynamic range, detection sensitivity, lower limit of detection (LOD), and signal-to-noise ratios (SNR) were examined. The relationships between exposure time, signal saturation, and linear dynamic range were also explored.

2. Experimental Design The goal of this study was to compare the limit of detection and linear dynamic range of imaging technologies commonly used for chemiluminescent Western blots. Two different CCD imaging systems and film exposure were chosen for testing.

To ensure unbiased data interpretation, we sought to minimize uncontrolled variables that commonly affect data acquisition. The most disruptive variable in chemiluminescent Western blotting is the kinetic, unstable nature of the detection chemistry. Chemiluminescent signals are governed by enzyme-substrate kinetics and are highly dependent on timing. To achieve maximum linear dynamic affected by mechanical shock and vibration, accidental exposure to light, temperature variation and aging. range, most detection methods require comparisonOther of factors multiple can affect the exposuresreading of luminometer tos suchdetermine as, lint, dirt or liquid the fumes optimal collected by the optic, liquid splashed on the optic, and mechanical or optical misalignment of the reading mechanism. 2 exposure time for each experiment. Because the Inintensity most application, of especially chemiluminescent in clinical applications, it is very signals important to verifychanges the performan ce of luminometers, because faulty readings can result in misdiagnosis of patients. over time, each exposure represents a unique stage in the kinetics of the enzyme/substrate reaction Good Laboratory Practice (GLPs) and many regulatory agencies, laws (such as CLIA 88) require the that cannot be recaptured. In this study, we removedluminometers this be variable periodically checked by tousing ensure that a the calibrated machine works as perluminometer factory specs.

reference plate (Harta Instruments, Fig. 1) as a stableQuite often light when thesource. reading of a certain samples are not correct, the lab personnel automatically blames the luminometer, and they ship the machine back to the manufacturer, when the problem is with the reagent, (bad reagent, wrong temperature, contamination, etc), the process of preparing the reagent, or any myriad of problems associated with anything but the machine. This causes a lot of unnecessary The Harta reference plate is a NIST-traceable standardexpenses andwidely interruption used of service, to when verify all that needsand to bevalidate done is to read the a reference sen plate- to verify the performance of the luminometer. 3 sitivity, linearity, and dynamic range of luminometers and other imaging instrumentation. Highly The Reference plate has dual redundancy light sources, so one light source can be compared against the stable LEDs and a linear optical attenuator systemother. generate The reference platemultiple has a built inlight battery cpointsheck system. acrAs longoss as the seven battery check orders light is ON, the batteries are OK. These are provided to check the performance of the reference plate itself. When the of magnitude. Reference wells that emit very low batterylevels goes belowof light a safe level, enable well A8 will evaluation turn off to indicate weakof sensitivitybattery. and

limit of detection (LOD). With the Harta device, weThe isolatedHarta RM-168-96 and luminometer examined reference plate the can beperformance used with any standard 96of well each microtiterplate luminometers. detection system in the absence of other experimental variables. These results reflect the intrinsic capabilities and limitations of each detection system,Since 1999, rather the Harta thanluminometer limitations reference plate h asof been the used Westernby thousands of satisfiedblot customers in over 25 countries worldwide. technique or detection chemistry.

  BOTTOM VIEW OF THE REFERENCE PLATE. 

Top view Bottom view

Figure 1. Luminometer reference microplate. The Harta RM-168 device uses stable LEDs and a linear optical attenuator

NISTsystem Traceable to generate Luminometer stable, Reference reproducible Microplate: light output across seven orders of magnitude. Light wavelength is 540 nm, and long- term stability is 5%. Features: • NIST traceable • CE compliant • 7 decades of stable light sources, checks: o Accuracy o Sensitivity o Dynamic Range o Linearity o Stability / repeatability -18 LI-COR Biosciences• Lowest level of light is equivalent to approximately 10 moles of luciferase using Promega www.licor.com/bio Bright-GloTM luciferase assay • Works in any standard 96 well luminometer • Dual redundancy light sources for the plate’s own performance verification • Built-in battery check plus weak battery indicator • Extended life lithium batteries • On-Off switch • Robust CNC machined aircraft grade aluminum construction • Supplied with padded hard storage case, spare battery pack, 2 switching tools, screwdriver and spare screws for the battery compartment • All electronics / optics system, no radioactive materials used in the plate

The performance of a luminometer is largely dependent on the performance of a very delicate Photomultiplier Tube (PMT). The PMT which is actually an old vacuum tube technology device, is easily Page 3 –Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

3. Materials and Methods A Harta RM168-96 luminometer reference microplate was used as the light source for all experi- ments (recommended recalibration date 1/15/15). The plate’s secondary, independent light source was used to verify light output and battery strength. The reference plate was imaged with X-ray film, a conventional CCD imager (identified here as Imager B), and an Odyssey® Fc dual-mode imager (LI-COR Biosciences). Signals were quantified and background adjustment was applied.

Film: The Harta plate was used to generate a series of film exposures. In a darkroom, the plate was placed face-down onto Blue Lite autorad film (GeneMate). For each exposure, the light source was activated for the indicated time. Film was developed by standard procedures. Film images were digitized, and densitometry was used to analyze signal intensities. Upper limit of detection was inferred by densitometry, as indicated by loss of linear response (plateau and saturation) for stronger signals.

Imager B: A conventional, commercially-available CCD imager was used. The Harta plate was placed face-up inside the imager, and activated. Various exposures were then captured by man- ual adjustment of image acquisition settings. Default exposure times did not produce acceptable results, so exposure times were manually optimized to maximize detection sensitivity and minimize signal saturation. For all images shown, the default setting of 3x3 binning was used, but other bin settings (1x1, 2x2, and 4x4) were also evaluated. Saturated pixels in the resulting images were high- lighted in cyan blue.

Odyssey Fc Imager: The reference plate was placed face-up inside the imager, and activated. Var- ious exposures were then captured, using default settings. No pixel binning was performed. Images were analyzed with Image Studio™ Software v. 4.0. This software highlights saturated pixels in cyan, but no saturation was observed.

Data analysis: Signals were quantified and background subtracted. Lower LOD was defined as (3 x standard deviation of background) + mean background (n = 12; background levels were sam- pled for 12 different areas of each image). Dynamic range was calculated as the ratio of maximum signal response (raw signal intensity of the strongest non-saturated well in each image) to the weakest possible signal (detector noise; defined as mean background of each image, n = 12). Signal-to- noise was calculated as the ratio of signal intensity from each well to mean background (n = 12).

4. Results and Analysis 4.1 Linear dynamic range, sensitivity, and limit of detection (LOD) This study illustrates the profound, but frequently unrecognized, effects of imaging methods on experimental results. The three imaging methods examined here show dramatic variations in performance when compared side by side. However, these limitations may not be readily apparent in everyday use of a single method.

Comparison of the imaging results highlights several key differences in performance of the methods tested (Fig. 2). As expected, film exposure displayed a very limited linear dynamic range of only 2-3 wells on the reference plate. Longer exposures enabled detection of fainter samples. However,

LI-COR Biosciences www.licor.com/bio Page 4 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

plateau and saturation of stronger signals made it impossible to detect and quantify both faint and strong signals in a single exposure (Fig. 2A). The CCD system, Imager B, outperformed film with a linear dynamic range of 3-5 wells. However, it too was limited by saturation of strong signals at longer exposures; longer exposure narrowed the dynamic range rather than extending it (Fig. 2B). The Odyssey® Fc system yielded distinctly different results. The linear dynamic range for this system was wider (6 wells). But the most striking difference was the stability of the dynamic range (Fig. 2C), which was constant across all exposure times (ranging from 30 sec to 10 min). Faint signals were detected even at short exposures, and the linear response of stronger signals was not sacrificed.

These data highlight the importance of experimentally determining the available linear dynamic range of a detection method, which may be very different from the theoretical linear range of the system. These results also demonstrate that detection sensitivity must be determined by calculating the lower LOD, rather than by simple visual inspection. Fig. 2 depicts several faint spots that are clearly visible to the eye, but fall below the calculated lower LOD cutoff (indicated in red).

The linear dynamic ranges of each method, and effects of exposure time on dynamic range, are examined in Figure 3. The graph in Fig. 3A clearly shows the limitations of film exposure. Film’s dynamic range was very limited, and strong signals could not be adequately documented because the increasing optical density (OD) of film quickly plateaued and reached saturation (seen in 5-min and 20-min exposures; red and blue lines, respectively). The dynamic range of film was unpredictable and varied widely with exposure time (see Fig. 5 for details). Because the quality and usability of film-based results are dependent on exposure time, a range of exposures must be collected for every experiment, in an effort to capture all necessary information. Although film is generally considered to be very sensitive, the detection sensitivity of film was inferior to digital imaging methods in this study (digital methods were able to detect two additional wells).

A) Film B) Imager B C) Odyssey Fc

Figure 2. Limit of detection (LOD) and linear dynamic range vary widely with different detection methods. For each method, 5 different exposure times are shown. Upper LOD for each exposure is marked with a blue line; lower LOD is indicated by a red line. A) For film exposures, upper LOD (plateau and saturation of strong signals) was inferred by densitometry (dashed blue lines; 5-min and 20-min exposures). B) For Imager B, saturated pixels (shown in cyan) were indicated by the manufacturer’s analysis software. C) For Odyssey Fc imaging, no saturated pixels were observed. For all examples, lower LOD = (3 x Std Dev of background) + mean background. For mean background, n = 12.

LI-COR Biosciences www.licor.com/bio Page 5 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

Imager B displayed an improved linear dynamic range of up to 5 wells on the reference plate (Fig. 3B). However, optimization of exposure time was again an important factor. The widest dynamic ranges were observed at very short exposures. In addition, small changes in exposure time had a large impact on dynamic range (discussed in more detail in Fig. 6). Saturation of strong signals was the limiting factor in dynamic range. Maximum dynamic range (5 wells) was obtained with an exposure time of 0.8 sec. Extending the exposure time slightly, to 1 sec, caused saturation of the strongest signal (see Fig. 6). Imager B’s detection sensitivity was clearly superior to film.

The widest linear dynamic range was observed with the Odyssey® Fc system, which detected 6 wells without signal saturation. The Odyssey Fc imager’s detection sensitivity exceeded that of film and Imager B, even for the shortest exposure time. Unlike the other imaging methods, signal intensities did not increase with longer exposures. Data from all exposure times were directly overlapping (Fig. 3C). This characteristic may seem counterintuitive at first, but it is the logical result of how image data are collected and expressed by this system (see below).

Data collection: number of photons vs. rate of photon collection Image data from chemiluminescent Western blots are typically collected in an additive manner, rep- resenting the total volume of photons collected from the sample. It logically follows, then, that longer exposures will allow collection of more photons. This increases signal intensity and moves the resulting data points upward on the y-axis. This is the case for both film exposure and conventional CCD imaging systems such as Imager B. With this method of data collection, image background is a critical limiting factor for detection sensitivity. During a long exposure, both signal intensity

A) Film B) Imager B

Figure 3. Exposure time affects performance, variability, and lower limit of detection. Quantification and corresponding images C) Odyssey Fc are shown for each detection method. Lower LOD = (3 x Std Dev of background) + mean background (n = 12). Boxes indicate signals within the linear dynamic range. A) Film exposure: Plateau and saturation of strong signals were inferred by densitometry (5 min, 20 min); indicated by dashed lines (red and blue, respectively). At 20 sec (green), dashed line indicates values below the lower LOD. B) Imager B: Dashed lines indicate values below the calculated lower LOD. Asterisks indicate saturated signals that could not be quantified. C) Odyssey Fc: Dashed lines indicate values below the calculated lower LOD.

LI-COR Biosciences www.licor.com/bio Page 6 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

and background levels (sometimes called the “noise floor”) will increase. Higher background may obscure faint signals, so a long exposure may not always improve the dynamic range and/or lower LOD.

The Odyssey® Fc imager is also a CCD system, but employs a different approach to data collection. It reports the rate at which photons are collected, rather than the total volume collected. The Harta luminescent reference plate emits photons at a constant rate, unaffected by exposure time. There- fore, when the number of photons collected per unit time is displayed graphically for various exposures, the resulting data points overlap on the y-axis (as seen in Fig. 3C). Because the rate of noise collection is also constant, and noise level is small relative to the photon collection rate, longer exposures do not increase the noise floor and will not obscure faint signals. When the rate of photon collection is measured, the results reflect the performance and capability of the overall system and experiment, rather than the performance and limitations of an individual exposure (Figs. 3A and B).

4.2 Experimental linear dynamic range is more limited than theoretical dynamic range Theoretical linear dynamic range is the maximum range that a device could reproduce if all other constraints are eliminated. It encompasses both extremes, from maximum response to the lowest possible signal intensity. In practice, however, constraints do exist and the full theoretical dynamic range is not accessible.

Detection of very faint signals is usually limited by noise. Faint signals must be strong enough to be distinguished from detector noise (the random noise that occurs in all electronic systems). If detector noise is very low, the lower limit of detection will improve and linear dynamic range will be extended. In this analysis, the width of the experimental linear dynamic range was calculated as the ratio of the strongest non-saturated signal to the weakest possible signal (detector noise; defined as the mean background of the image).

Film and densitometry. When the linear dynamic range of an imaging method or system is de- scribed, the theoretical value is typically used. For film, the dynamic range is considered to be 1-1.5 orders of magnitude.2,4 That range is dictated by the photochemistry of film exposure. Photographic emulsions, first introduced in the 1850s, record light emission by activation of silver halide crystals (or grains) on the film surface. Photons of light activate the silver grains by converting silver ions to silver atoms, to create a latent image. Film processing uses these silver atoms to catalyze reduction of the entire activated grain to black metallic silver, converting the latent image into a visible image. Unactivated silver crystals are washed away during processing. The concentration of metallic silver on the developed film is called optical density (OD).4,5

Although OD is related to the amount of light absorbed by photographic emulsion, it is not a direct function of light exposure. Film response is dependent on light intensity and duration of exposure; this relationship is called the Reciprocity Law. It is an inverse relationship: bright light delivered for a short duration can produce the same response as dim light delivered for a longer time. However, this relationship only holds true across a limited range of values. Outside that range, film response is not proportional to light intensity and exposure time. This effect, called “reciprocity failure”, occurs at both low and high intensities of light.6 Because of reciprocity failure, the response of film to light is logarithmic rather than linear.5 Faint and strong signals are both significantly under-represented, because film is inherently less responsive at these extremes.

LI-COR Biosciences www.licor.com/bio Page 7 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

Densitometry of exposed films is often used to quantify protein levels on Western blots. Densi- A) Film tometry measures the OD, or degree of darkness, as a function of the transmission of incident light through the developed film. Dark areas on the film transmit less light, and have a higher OD. The usable range of densities is approximately 0.2–2.0. The OD of unexposed film is approximately 0.2, and signals must exceed this density to be recorded. A medium-gray tone, halfway between white and black, reflects an OD of 1.0.7 High ODs, above 2.0, appear black and indicate signal saturation; differences in intensity cannot be discriminated.

Exposed films must be digitized for densitometry. B) Imager B The scanning process is also a significant source of error and can compromise data integrity. Desktop scanners often do not accurately reproduce the original data, and the dynamic range of scanned images is lower than that of CCD images.8 Many scanners truncate large signal peaks to limit the digital output to a factory pre-set range. This type of default image adjustment is performed automatically by the scanner, and the user is not notified that data truncation has occurred.

In this study, the observed linear dynamic range C) Odyssey® Fc of film was extremely limited. A 1-min exposure yielded the most optimal result, with a dynamic range of 1.25-fold, represented by three wells on the reference plate (Fig. 4A). Across this range, the R2 value for a linear regression was only 0.716. Saturation of strong signals was an important limiting factor in dynamic range (shown in more detail in Figs. 5A and B), and resulted in a much lower range than the theoretical capacity of 10- fold. This observation is consistent with published reports.9 Figure 4. Linear dynamic range (LDR) varied dramatically. The exposure with the widest linear dynamic range was identified for CCD imaging with Imager B. For this CCD sys- each method. LDR = ratio of maximum signal response (strongest tem, the manufacturer’s technical specifications non-saturated well) to weakest possible signal (mean background state a theoretical linear dynamic range of > 4 of image). Boxes indicate signals within the LDR. Dashed lines logs. This figure reflects the bit-depth of the elec- (red) indicate values below the lower LOD. For Imager B, the tronics (16 bit, corresponding to 65,535 shades of default 3x3 bin setting was used. grey). In this study, the optimal exposure time was 0.8 sec (Fig. 4B), with a dynamic range of 198- fold, represented by five wells on the reference plate. Across that range, the R2 value for a linear

LI-COR Biosciences www.licor.com/bio Page 8 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

regression was 0.9153. Exposures longer than 0.8 sec improved the lower LOD but also caused saturation of strong signals (Fig. 3B), leading to no overall improvement in dynamic range. As with film, signal saturation was clearly a limiting factor in dynamic range. The experimental range was much lower than the > 4 log capacity of the imaging hardware. Similar reductions in actual performance have been reported by others.2,10-11

The bin settings chosen for image capture can significantly affect signal intensity and linear dynamic range. Binning combines the photoelectric charges from adjacent pixels in a CCD image sensor, prior to signal readout.12 The default bin setting for Imager B was 3x3, but 4x4, 2x2, and 1x1 binning were also examined. Higher bin settings, such as 4x4, increase signal-to-noise ratio because signals from multiple pixels are combined during readout. Although this improves detection of faint signals, it also forces saturation of strong signals and may narrow the linear dynamic range. Binning inherently sacrifices spatial resolution by reducing the number of pixels in the image. 2x2 binning lowers signal intensities, relieving the saturation of strong signals but also failing to detect fainter signals.

CCD imaging with Odyssey® Fc system. The manufacturer’s specifications indicate a theoretical linear dynamic range of > 6 logs for this system. This study displayed a range of 8,617-fold, represented by six wells on the reference plate (Fig. 4C). Across that range, the R2 value for a linear regression was 0.9387. Exposure time was not a factor in dynamic range or lower LOD; all exposure times tested yielded the same result (Fig. 3C). For the comparison in Figure 4, the 10-min exposure was chosen. Because the strongest signals did not saturate the detector, a very wide linear dynamic range was captured in a single exposure. Although the same lower LOD could be achieved with Imager B and the Odyssey Fc system, Imager B required an exposure time that saturated strong signals and reduced the available linear dynamic range to only 3 wells (Figs. 2B, 3B). The experimental range reported here for the Odyssey Fc is lower than the theoretical capacity of > 6 logs for this imager.

Limitations of the Harta reference plate. The observed dynamic range and lower LOD of all three methods were likely affected by an unavoidable limitation of this experimental system. The luminescent intensities of the Harta plate span an extremely wide range across only eight wells. This necessitates large changes in light intensity from one well to the next. If the actual lower LOD falls in the large interval between wells, both linear dynamic range and lower LOD would be underestimated. This could have an especially large impact on film detection, because reciprocity failure will cause under-representation of both low- and high-intensity signals.6 If light intensity was reduced by smaller intervals, all detection methods would likely have achieved a somewhat wider linear dynamic range. Unfortunately, an ideal calibrated light source is not commercially available.

4.3 Impact of exposure time on linear dynamic range Film and photochemistry. Exposure time must typically be optimized to achieve the maximum linear dynamic range. This can be a time-consuming and iterative process, because the dynamic range is unpredictable and even small changes in exposure time may shift the dynamic range upward or downward.

With film exposure, the interplay between exposure time and dynamic range is a major disadvan- tage. When signals are very intense, high-intensity reciprocity failure occurs and film response becomes logarithmic rather than linear.5 When all silver grains have been activated, film response is

LI-COR Biosciences www.licor.com/bio Page 9 –Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

saturated and no further signal can be recorded. But film response also plateaus ellw before saturation is reached, due to the statistics of silver grain activation. As film is exposed to high-intensity light, more silver grains are activated and each new photon becomes statistically less likely to strike an unactivated grain.4,6 Strong signals are therefore under-represented on film, and a logarithmic response in OD occurs prior to saturation. These signals cannot produce the appropriate tonal variations, causing moderate-intensity and high-intensity signals to appear similarly dark.

Linear range is severely limited by high-intensity reciprocity failure.6 To make matters worse, there is no clear indication that reciprocity failure has occurred. CCD imagers typically designate saturated pixels with a contrasting color such as cyan or red, but film exposure offers no clear indication of plateau or saturation. The graph in Figure 5A demonstrates reciprocity failure in long film exposures. In a 5-min exposure (red line), film response has flattened significantly and brighter wells produce little increase in OD. After 20 min, saturation is even more pronounced (blue line). Higher OD values should be observed for the extended 20-min exposure, but these signals have reached saturation and are no higher than the 5-min values. Figure 5B shows that signal plateau has begun to occur even after 1 min of exposure (purple line).

Imager B. Small changes in exposure time can modify and shift the linear dynamic range, even when a CCD imaging system is used. Figure 6A shows three different exposure times that varied only slightly (0.5 sec, 0.8 sec, and 1 sec). Although the exposures only differ by a few tenths of a second, both the upper and lower limits of detection were affected. At 0.5 sec (green line), the dynamic range spanned four wells. At 0.8 sec (red line), an additional well was visualized and the dynamic range expanded to five wells. However, at 1.0 sec (blue line), the dynamic range narrowed again, to only four wells. The boundaries of the dynamic range also shifted along the x-axis, as the strongest well saturated the detection (indicated by blue pixels).

The performance of this imaging system treads a very fine line, because strong signals are so easily saturated. Here, two-tenths of a second pushed a strong signal past the capacity of the detector. Longer exposures are required to detect fainter wells, but cause saturation at the upper end. Even with careful optimization of image acquisition settings, the full range of faint and strong signals could not be captured in a single exposure.

A) B)

Figure 5. With film, stronger signals are under-represented because they plateau and become saturated. Signal intensities were determined by densitometry, and background subtraction performed. Signals that fall within the linear dynamic range are boxed in each image. A) For longer exposures, the saturation effect is very pronounced. For the brightest wells, a 20-min film exposure (blue line) cannot record a higher optical density than a 5-min exposure (red line). B) Signal even begins to plateau after a 1-min exposure (purple line).

LI-COR Biosciences www.licor.com/bio Page 10 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

Multiple exposures. With both film exposure and Imager B, multiple exposures were needed to capture and define their linear dynamic ranges. In this study, the Harta reference plate provided a constant, reproducible light source to facilitate the capture and direct comparison of many different exposures. However, a chemiluminescent Western blot presents additional challenges. The chemiluminescent reaction is dynamic, and does not generate stable, constant levels of light. When multiple exposures are captured and compared, each exposure represents a unique context – making it more difficult to choose appropriate exposure times, compare results, and capture the full range of data from the blot.2

Odyssey® Fc. The performance of this CCD system was fundamentally different from Imager B. Linear dynamic range was unaffected by exposure time, and remained constant for exposures ranging from 30 sec to 10 min (Figure 6B). Because this system accommodated a wide range of signal intensities without saturation, it was possible to capture all necessary information in a single image acquisition. The ability to capture the full range of data in a single exposure simplifies data analysis. A single acquisition also enhances the consistency and reproducibility of data capture in the dynamic, enzymatic context of a chemiluminescent Western blot.

4.4 Signal-to-Noise Analysis Signal-to-noise ratio (SNR) is where limit of detection and linear dynamic range come together to determine the overall performance of an imaging method (Table I). If the dynamic range is narrow, SNR will be poor. The signal-to-noise relationship of the strongest signal in the image dictates the maximum linear dynamic range, and SNR for all other data points will be lower than this value. Film exposure (Fig. 7A; Table I) displayed very low SNR values that varied dramatically with exposure time. Maximum SNR was only 25% above background. With Imager B, SNR was much improved but was still highly dependent on exposure time (Fig. 7B; Table I). For longer exposures, the strongest wells frequently yielded saturation and these values could not be plotted (indicated by asterisks in Fig. 7). The Odyssey Fc system produced SNR values much higher than those of other methods tested (Fig. 7C; Table I), and SNR was unaffected by exposure time. Figure 8 shows a comparison of the optimal SNR data collected with all three imaging methods.

A) Imager B B) Odyssey Fc

Figure 6. Small changes in exposure time can shift the linear dynamic range. A) Data from Imager B are shown for three exposure times. The linear dynamic range was variable, and affected by very small changes in exposure time (0.2-0.3 sec). Boxes indicate signals within the dynamic range. Cyan pixels (1-sec image) indicate saturation of the detector and loss of linear dynamic range. B) Data from Odyssey Fc imager, for five exposure times. Linear dynamic range and lower LOD were constant for all exposure times. No saturated pixels were observed in any exposures.

LI-COR Biosciences www.licor.com/bio Page 11 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

Table 1. Performance of the three detection methods. Values shown represent the maximum for each method.

Method Lower LOD LDR (wells) LDR R2 across SNR SNR LDR (max) (min)

Film 4th well 3 wells 1.25-fold 0.716 1.248 1.099 Imager B 6th well 5 wells 198-fold 0.9153 197.700 1.259 Odyssey® Fc 6th well 6 wells 8,617-fold 0.9387 8616.505 9.709

A) Film: Signal-to-noise ratio

B) Imager B: Signal-to-noise ratio C) Odyssey Fc: Signal-to-noise ratio

Figure 7. Signal-to-noise ratios (SNR) are boosted by wide linear dynamic range. SNR was calculated for each detection method, for multiple exposure times. SNR values <1 were considered equivalent to background, and are not shown. * indicates saturated signals that could not be accurately calculated.

LI-COR Biosciences www.licor.com/bio Page 12 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

n Film, 1 min n Imager B, 0.8 sec n Odyssey® Fc, 10 min

Figure 8. Signal-to-noise ratios highlight the dramatic differences between the three methods tested. The exposure time with the broadest linear dynamic range was selected for each method (as in Figure 4). SNR was calculated and plotted.

5. Discussion Choice of imaging method can have unintended effects on Western blotting results, but the magnitude of these effects may be unrecognized or underestimated. Side-by-side comparison of three methods in this study revealed dramatic variations in performance and linear dynamic range (Table I). However, the limitations of each method may be much less apparent when a single method is used without comparison.

5.1 Choosing an appropriate detection method This study demonstrates the importance of choosing an appropriate Western blot detection method for the type of data analysis that will be performed. Although detection sensitivity is often a high priority, linear dynamic range may be equally important. It is important to be aware of the strengths and limitations of available detection methods. Matching the detection method to the experimental goals will help to ensure that the appropriate detail and complexity of the data are captured in each experiment.

Here, a calibrated light source was used to eliminate possible sources of error and variability. But at the bench, the Western blotting process itself contributes additional sources of error. Chemilu- minescent detection is a dynamic process, subject to enzyme/substrate kinetics. Unlike the Harta reference plate, the chemiluminescent reaction does not generate constant light output. While the researcher is taking multiple exposures of the blot, trying to capture the optimal dynamic range, signal intensities may be changing. Because small changes in exposure time can have considerable impact, the dynamic context of a chemiluminescent Western blot makes it even more difficult to compare images and identify the most appropriate exposure time to capture the full range of data.

LI-COR Biosciences www.licor.com/bio Page 13 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

Data capture and analysis are more reproducible and straightforward when a single image acquisition provides the maximum linear dynamic range, with excellent detection sensitivity and without saturation of strong signals. Film and Imager B were unable to meet that requirement in this study. However, the Odyssey® Fc system did provide consistent detection sensitivity and linear dynamic range for all exposure times evaluated.

5.2 Impact of signal saturation and lower limit of detection This study explored several key factors that affect the linear dynamic range of detection. Saturation of strong signals was a significant limitation for both film and Imager B. For both methods, it was critical to analyze and compare multiple exposures to minimize the impact of saturation. Very small changes in exposure time had a considerable effect on both signal saturation and linear dynamic range. Exposure time may have an even greater impact on dynamic range when the time-dependent chemistry of chemiluminescent detection is introduced as another variable.

Detection sensitivity and dynamic range also go hand-in-hand. If the lower LOD is poor, dynamic range will be limited. The ability to detect faint signals above background, combined with detection of strong signals without saturation, enables a wider linear dynamic range.

Film exposure is constrained by the photochemistry of detection. This study demonstrates that detection of strong signals is severely limited by high-intensity reciprocity failure – which, in turn, greatly limits linear dynamic range. To make matters worse, it is difficult to identify the point at which reciprocity failure begins to affect linear response and data output, and every experiment will be different. Unlike CCD imagers, which use contrasting colors to indicate saturated pixels, film offers no visual indication of signal plateau or saturation.

For digital imaging, performance is linked to the system’s optical technology and engineering. Detection sensitivity and lower LOD are ultimately determined by system noise. Different optical systems use different approaches to mitigate noise limitation, and an optical system with low detector noise may offer enhanced detection sensitivity. If the optical system can simultaneously provide low detector noise and avoid saturation of strong signals, very wide linear dynamic range can be achieved. This study illustrates considerable variations in performance between Imager B and the Odyssey Fc system, two differently-engineered CCD imaging systems.

Although technologies such as near-infrared fluorescent detection can eliminate the time-dependent dynamics of chemiluminescent Western blotting, some researchers may prefer to continue using established chemiluminescent methods. An imaging method with higher performance can eliminate other key variables, such as exposure time, from the detection process. This enhances the accuracy and reproducibility of chemiluminescent detection, because all data is captured in a single exposure.

5.3 Enhancing Western blot reproducibility Western blotting is as much an art as it is a science. Subjective factors, such as exposure time, greatly affect performance and outcome. A less-than-optimal exposure is unlikely to yield all the

LI-COR Biosciences www.licor.com/bio Page 14 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

information needed for accurate data interpretation – but optimal exposure time is difficult to determine, and is different for every experiment. Recently, reproducibility of Western blot data has become a topic of increased interest.13-14 Reproducibility is enhanced when subjective factors and inherent sources of variability are reduced or eliminated. Wide linear dynamic range makes it possible to remove exposure time as a variable. This reduces the subjectivity of Western blot analysis and makes data output more consistent, making Western blotting less of an art and more of a science.

LI-COR Biosciences www.licor.com/bio Page 15 – Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

6. References 1. Mitchell, P. Proteomics retrenches. Nature Biotechnol. 28: 665-670 (2010).

2. Mathews ST, EP Plaisance, and T Kim. Imaging systems for Westerns: chemiluminescence vs. infrared detection. Methods Mol Biol. 536: 499-513 (2009).

3. Ozaki DA, H Gao, CA Todd, KM Greene, DC Montefiori, and M Sarzotti-Kelsoeet. Internation- al technology transfer of a GCLP-compliant HIV-1 neutralizing antibody assay for human clinical trials. PLoS ONE 7(1): e30963 (2012).

4. Baskin DG and WL Stahl. Fundamentals of quantitative autoradiography by computer densitometry for in situ hybridization, with emphasis on 33P. J Histochemistry Cytochemistry 41(12):1767-76 (1993).

5. Kodak. The essential reference guide for filmmakers: Basic sensitometry and characteristics of film. Kodak Educational Products, Code: H-845 CAT No. 145 6144. Eastman Kodak Com- pany (2007).

6. Laskey RA. Methods of detecting biomolecules by autoradiography, fluorography and chemiluminescence. Amersham Life Sci Review 23: Part II (1993).

7. Davis R and FM Walters Jr. Sensitometry of photographic emulsions and a survey of the characteristics of plates and films of American manufacture. Scientific Papers Bureau of Standards 18: 1-120 (1922).

8. Gassmann M, B Grenacher, B Rohde, and J Vogel. Quantifying Western blots: pitfalls of densitometry. Electrophoresis 30: 1845-55 (2009).

9. Wang YV, M Wad, ET Wong, YC Li, LW Rodewald, and GM Wahl. Quantitative analysis reveal the importance of regulated Hdmx degradation for P53 activation. Proc Natl Acad Sci USA 104(30): 12365-12370 (2007).

10. Galownia NC, K Kushiro, Y Gong, and AR Asthagiri. Selective desensitization of growth factor signaling by cell adhesion to fibronectin. J Biol Chem 282: 21758-21766 (2007).

11. Huang D and SA Amero. Measurement of antigen by enhanced chemiluminescent Western blot. BioTechniques 22: 454-458 (1997).

12. Farrell J, M Okincha, M Parmar, and B Wandell. Using visible SNR (vSNR) to compare the image quality of pixel binning and digital resizing. Proc SPIE 7537: 75370C (2010). doi:10.1117/12.839149

13. Editorial. A picture worth a thousand words (of explanation). Nature Meth 3(4): 237 (2006).

14. Collins FS and LA Tabak. NIH plans to enhance reproducibility. Nature 505: 612-13 (2014).

LI-COR Biosciences www.licor.com/bio Page 16 –Film and CCD Imaging of Western Blots: Exposure time, signal saturation, and linear dynamic range

LI-COR is an ISO 9001 registered company. © 2014 LI-COR, Inc. LI-COR and Odyssey are trademarks or registered trademarks of LI-COR, Inc. in the United States and other countries. All other trademarks belong to their respective owners. For patent information, visit www.licor. com/patents.

Global Headquarters, Serving the United States 4647 Superior Street • Lincoln, NE 68504 • Phone: +1-402-467-0700 • Toll free: 800-645-4267 • Fax: +1-402-467-0819 [email protected]

Regional Offices LI-COR GmbH, Germany, Serving Europe, Africa, and the Middle East. LI-COR Biosciences GmbH • Siemensstraße 25A • 61352 Bad Homburg • Germany • Phone: +49 (0) 6172 17 17 771 Fax: +49 (0) 6172 17 17 799 • [email protected]

LI-COR Ltd., United Kingdom, Serving Denmark, Finland, Iceland, Ireland, Norway, Sweden, and UK. LI-COR Biosciences UK Ltd. • St. John’s Innovation Centre • Cowley Road • Cambridge • CB4 0WS • United Kingdom Phone: +44 (0) 1223 422104 • Fax: +44 (0) 1223 422105 • [email protected]

LI-COR Distributor Network: www.licor.com/bio/distributors

Doc # 979-15033 11/14

LI-COR Biosciences www.licor.com/bio