For most analytical techniques, the key to obtain accurate and reproducible results is in understanding the limits of the tools employed and defining the lower and upper limits of quantitation and the linear dynamic range. Measurements using standard curves remain the best method for accurately defining these parameters, and are typically performed using serial dilutions of a representative sample [8, 9]. The requirement for clean and interpretable data is critical for determining these two parameters. Western blotting involves the following complex series of steps and obtaining quantifiable results requires that all the steps be performed rigorously as most of them are interdependent:
-
1.
Lysis of cells or tissue
-
2.
Quantification of lysate total protein concentration
-
3.
Equal loading and separation of samples using SDS–PAGE
-
4.
Complete transfer of proteins separated on the gel to nitrocellulose or PVDF membrane
-
5.
Determination of proper dilutions for primary and secondary antibodies
-
6.
Optimal detection of the chemiluminescent signal
-
7.
Quantification of densitometric data
Thus, a methodology to test and confirm the quality of each step should be employed.
Validation of the Loaded and Transferred Proteins
The handling, storage, and lysate preparation of tissue and cell culture specimens can be complicated by the nature of the sample itself and by the large array of reagents and equipment available to prepare them. Careful consideration should be given to determine the best approach to prepare samples in order to get a reliable end result. Since most lysis buffers contain detergents such as Triton X-100 or sodium dodecyl sulphate (SDS), a detergent-compatible protein assay should be chosen.
The typical amount of lysate loaded per lane of an SDS–PAGE gel is between 10 and 80 μg, often times, with the same amount of protein per lane loaded regardless of the antibody used or target probed. It is also a common practice to run the gel and then immediately transfer the separated proteins to PVDF or nitrocellulose membrane without verifying the in-gel consistency of protein loading or of the quality of separation. In this study, the stain-free gel technology (Bio-Rad) was used to image the SDS–PAGE separation of a two-fold dilution series of a HeLa cell lysate prior to transfer (Fig. 2a). The quantitative analysis of the total protein loaded per lane gave an excellent correlation (R
2 = 0.9855) to the two-fold dilutions over a linear dynamic range between about 1 and 35 μg of the loaded protein (Fig. 2b).
Following the transfer, the membranes were imaged to assess the transfer efficiency and lane-to-lane consistency of the transferred proteins, which retain the fluorescence imparted by the stain-free imaging process (Fig. 2c). Although, the linear dynamic range was much narrower for the transferred protein, a very good correlation (R
2 = 0.9971) to the two-fold dilutions was obtained in the typical range (between about 10 and 70 μg) of total protein load for most western blotting techniques (Fig. 2d). The reduced sensitivity of the fluorescent signal from the membrane versus the gel is the result of some protein transferring through the membrane (over-transfer) and much higher background due to the autofluorescence of the PVDF membrane under UV illumination during stain-free imaging. The combined effects of over-transfer and variable transfer efficiency support the use of densitometric data produced from the membrane for normalization as opposed to the gel.
The Quantitative Linear Dynamic Range: Contrasting Detection Platforms
Film has been the traditional method of choice for the detection of chemiluminescent western blots using a wide variety of detection reagents and horseradish peroxidase (HRP)-conjugated secondary antibodies. Although, film provides excellent resolution and sensitivity, the dynamic range of quantitation is poor [7]. On the other hand, with the next generation camera-based detection methods, both the sensitivity and linear dynamic range are excellent, which permits much more accurate quantification of the relative density between samples. This is illustrated in the comparison of image data for a two-fold dilution series of ADH generated from the same blot using film and camera-based detection methods (Fig. 3a, b). Here, a linear dynamic range of four dilutions (16-fold) between 0.04 and 0.31 ng was observed for ADH with film as opposed to the seven dilutions (128-fold) between 0.04 and 2.5 ng for the ChemiDoc MP imager (Fig. 4).
For the endogenous protein GAPDH, both film and the ChemiDoc MP were only linear for the lowest three dilutions. This can be attributed to very high abundance of this protein in the HeLa cell lysate (Fig. 3). In this case, more dilutions and a longer exposure time would extend the linear dynamic range, but these data show that to use GAPDH as a reliable and quantifiable loading control, a total protein load of not more than about 0.5 μg is required (Fig. 3).
For the film images, a slight decline in the density was observed with increasing protein load within the plateau region (Fig. 3b). This can be attributed to the saturation of background between and around the bands of interest. Given that the ADH and GAPDH bands of interest are already fully saturated at 4.4 μg of total protein load, the contribution of the increasing background density (that saturates quickly with the film) results in a net decrease in background-subtracted density for these proteins.
Optimization of Protein Loading for Quantitation
All blot and gel detection systems generate a two-dimensional image, and signal intensity forms the third dimension of information (Fig. 1). Therefore, if excessive protein is loaded in the gel lane such that the width of the gel has been filled, the detector (whether film, camera, or scanner) will only capture the signal from the protein that resides near the surface of the gel. The same effect is observed during blotting where the protein transferred from an overloaded gel lane will form a layer on top of the protein already bound to the surface of the transfer membrane such that the primary antibodies will only bind to the surface layer of the transferred protein [10, 11]. The plateau observed for the same blot using both film- and camera-based detection is not due to the saturation of the detector, but rather is the result of protein saturation within the gel itself and consequently of the blot itself with the multiple layers of target protein bound to the membrane (Fig. 3).
Since many labs typically load a specific, fixed amount of protein (typically between 10 and 50 μg) into the gel lanes without determining the optimal protein load, there is a strong potential for gel saturation, particularly for loading controls such as GAPDH, beta-actin, and tubulin. Although, this results in consistent band densities among the sample lanes, these data are often the result of overloading the gel for these abundant proteins such that the densities obtained are far outside the linear dynamic range in the plateau. The best way to avoid this issue is to produce a standard curve from a two-fold serial dilution series over 12 dilutions starting from about 80 μg of a pooled lysate from representative samples across the experimental conditions [8]. A separate standard curve of band density versus protein load should be run to validate each primary antibody for western blot as shown in Fig. 3. In this fashion, the linear dynamic range of detection for each antibody can be determined, and the associated dilution factor required for individual sample loading can be obtained by diluting the samples to the mid point of the standard curve. In the present case, we determined that the linear dynamic range of ADH is between 0.04 and 2.5 ng of the purified protein (Fig. 4b, c).
Loading Controls for Quantitation
The transfer efficiency of protein to a blotting membrane can be inconsistent across the gel, resulting in a gradual two- to four-fold increase or decrease in the signal between the lanes. Furthermore, the preparation and quantitation of sample lysates for the concentrations coupled with their physical pipetting into the lanes of a protein gel can also lead to inconsistent densitometric data. Loading controls are useful to normalize these technical artifacts and become increasingly important when measuring small differences in protein expression between samples.
The most common loading controls include housekeeping proteins, such as GAPDH, beta-actin and tubulin, which are constitutively expressed proteins that maintain cell viability. However, these proteins are generally highly expressed in samples and are frequently overloaded in the gel lane with the target protein such that they would not serve to normalize the loading [12, 13]. Although, the densitometric data would be consistent, this data could be an artifact of overloading as observed in our experiments where GAPDH levels were reaching a plateau above 0.5 μg of HeLa lysate (Fig. 3). Furthermore, the housekeeping proteins themselves can be variably expressed between the experimental conditions, thereby eliminating their usefulness for the normalization of western blots [14–17]. In order to avoid these issues, the total lane density of transferred protein on the membrane is being used for the normalization in lieu of the traditional loading controls [18–22].
Since the stain-free gel technology produced accurate and quantitative densitometric data from the captured fluorescence intensity of the transferred protein on the blot (in the range of 10–70 μg of HeLa lysate protein loaded per lane; Fig. 2), we tested the correlation of lane density with actual protein load. HeLa cell lysate samples of known concentration with spiked-in ADH were separated using stain-free gels and then imaged to verify uniform loading and separation prior to blotting (Fig. 5a, inset). The proteins were then transferred to PVDF membranes and the blots were imaged using the stain-free imaging application on the ChemiDoc MP prior to antibody incubation to validate transfer efficiency, and to assure complete protein transfer to the membrane (Fig. 5a). We then compared the relative total lane density from the stain-free blot image with the relative μg quantity of HeLa lysate protein load (Fig. 5a), and found a positive Pearson Correlation (p value of 0.0398) supporting this method as a valid approach to normalize western blot data.
The relative intensities of the GAPDH bands were approximately equal among the lane groups 1–4 for both image-based (Fig. 6) and film-based (Fig. 7) detection. This data did not correlate well with the two-fold relative difference in μg quantity of HeLa lysate protein load yielding a negative Pearson Correlation (Figs. 6, 7, panels B and D). In contrast, the total lane density of transferred protein on the blots produced a better correlation with the fold change in protein load for the same lane groups (1–4), with a positive Pearson Correlation (p value of 0.0398) (Fig. 5b). This can be explained by contrasting the linear dynamic range for GAPDH and total protein, where at protein loads above about 0.5 μg, GAPDH is saturated (Fig. 3), and in the plateau whereas total protein lane density on the transferred blot is within the linear range of 10–70 μg (Fig. 2d). For lane groups 1–4, the protein loads between 11 and 22 μg, were well above the linear dynamic range of GAPDH but within that of the stain-free detection explaining much better quantification for the latter.
The HeLa lysate total protein load of 0.34 μg (lane groups 5 and 6) was well below the dynamic range and even below the detectable limit for stain-free imaging (Fig. 5a), but within the linear dynamic range for GAPDH (Fig. 3). This makes GAPDH an ideal loading control for highly abundant proteins which require much lower sample loading. It is worth pointing out that in these cases, stain-free imaging cannot be used for normalization. Care must be taken to ensure that the amount of lysate loaded in this case is within the linear dynamic range of both the loading control and the target protein to ensure accurate, quantifiable, and normalized densitometric data.
Accurate Quantitation Using the Linear Dynamic Range
The 0.03-fold difference in HeLa lysate loading among lane groups 2, 5, and 6 (i.e., between 11 and 0.34 μg) was calculated to be only about 0.20–0.26-fold by relative band density of GAPDH (Figs. 6, 7, panel D contrast lane group 2 with 5 and 6 for “Fold GAPDH”). This can be explained by the fact that the total protein load of 0.34 μg is within the linear dynamic range of density for GAPDH but is in the plateau at 11 μg (Fig. 3). Thus, the densitometric fold difference is much smaller than expected because the two points are not within the linear range. This further underscores the importance of ensuring that the samples are diluted such that the loading control is well within the linear dynamic range of detection.
For ADH, the correlation between relative density and fold difference in ng quantity of spiked-in ADH protein load between all the lane groups was excellent when using the ChemiDoc MP imager (Fig. 6c), but poor with film (Fig. 7c). This is due to the different linear dynamic ranges obtained between film and the ChemiDoc MP for ADH (Fig. 4). All the loaded amounts of ADH (i.e., 0.17–1.39 ng) were within the linear dynamic range of detection for the ChemiDoc MP (Fig. 4b). However, this was not the case for the film, where the relative densities for ADH were measured from one value within the linear dynamic range (0.17 ng) and two values within the plateau (0.69 and 1.39 ng) (Figs. 3, 4).