Autoradiogram image acquisition with a document scanner
Many scientists are now using standard commercial document scanners to capture images of electrophoresis gels, blots and autoradiograms with a view to further analysis. How valid are such scanners as image capture devices, given the difficulties and pitfalls of densitometry in general? Clearly the use of commercial scanners to analyse gels with respect to positional information such as Rf, pI and Molecular size determination poses no conceptual difficulty. However, there is some concern related to the optical range of document scanners and the linearity of data translation during the image formation.
Over the last few decades the traditional lane-scanning densitometers were best described as optically efficient, but time-consuming and awkward in their capture and analysis of gel data. Now that imaging laser densitometers and desktop scanners are being used, the highly efficient methods of image analysis can be employed to make data handling much faster, more controlled and much more flexible. For some densitometry applications, there is no doubt that high optical capabilities are required to achieve the desired experimental optical density range. However this can be costly and is unnecessary for many electrophoresis applications. Furthermore the best equipment is often not supplied with the best analysis software which is now just as important as the equipment. So, can an "off the shelf' scanner, combined with good quality software, become a suitable alternative to a purpose-built densitometer?
Scanner and Materials
In order to examine this question a standard Epson GT8000 scanner was used to scan several autoradiograms. The scanner was not specially optimised in any way. The autoradiograms were kindly donated by Dr L Field of Rothamsted Experimental Station and were produced using a 32P labelled probe that was hybridised to Southern blot filters containing several electrophoresed lanes of restricted total cell DNA. Fuji RX film was used for autoradiography without any pre-flashing, or pre-sensitising. In each gel the lanes were loaded with a doubling series of sample prior to electrophoresis. The captured images were then analysed using Phoretix 1 D Gel Analysis software.
One of the images, Gel 1, is shown in Figure 1 where the lanes have been identified using the analysis software. As an example the profile in Figure 2 is of lane 5 and is an averaged profile of the grey levels of the pixel points of the lane image below the graph. Densitometry measurements are performed in the software by summing all the pixel grey level values, after background subtraction, between the boundaries of the bands denoted by the dotted lines.
The Phoretix software allows automatic band peak detection and automatic band edge detection, both of which may be followed by simple, interactive and precise editing. Background subtraction was performed using a "Rubber Band" algorithm which acts by simulating the stretching of a rubber band beneath the averaged profile. This and other background subtraction techniques are automatic procedures within Phoretix software. Therefore each lane was treated independently through an automatic analysis in the software. The manual override facility for background subtraction was not found to be necessary for any of the lanes, or bands examined.
Scans 1 and 2 of Gel 1 were then re-analysed with the measurements being related to an intensity calibration standard curve produced from a scan of the step tablet at the same scan intensity used for scanning the respective autoradiogram. The data is shown in Figure 7 and clearly demonstrates that the intensity calibration is effective in normalising data from scans done at different scan intensifies. The close fit of the curves shown in Figure 7 is even more impressive when it is realised that each of the analyses were done quite separately which means that relative margin of error during the analysis was small. However, one problem that is apparent is the fact that the graphs do not reflect the slope that would be expected from a doubling series.
Figures 5, 6 and 7 show that the intensity calibration method for adjusting the nonlinearity of image production by the uncalibrated scanner is valid within the range of .05 to 1.4 optical density units. All the bands examined are within this range. We therefore need to examine other possibilities for the lack of correspondence to a doubling series. The three main possibilities are as follows:-
1 .Sample loading errors. 2. Non linearity of x-ray film darkening response. This possibility was considered especially likely because some of the bands are very dark and it is known that the film was not pre-sensitised. 3. Probe hybridisation dynamics. It is possible that the degree of probe binding is affected in a non linear fashion by the concentration of DNA present. This is more likely to be true for the weakly binding bands.
The third possibility is demonstrated by comparing the weakly binding bands in Gel 1 with the strongly binding band 7 in each lane. Within the data presented in Table 1 there are a number of bands in lanes 5, 6 and 7 that are similar in their measurement range as the the seventh band across all the lanes. Nevertheless, it is clear from Figure 8 that the rates of increase of the bands 1 to 6 have a completely different profile to that of band 7. This would suggest that there is a larger than expected effect of DNA concentration on probe hybridisation to those bands for which it is less specific.
Table 1.- Calibrated Data from Gel 1
In order to examine whether the rate of increase of band 7 in the lanes of Gel 1 was an effect of autoradiography exposure, several different autoradiograms were examined. In each case there was a doubling series of samples and there existed a wide range of exposure levels between the gels, as shown in Table 2. All of these autoradiograms were examined following intensity calibration as described above.
Table 2: Calibrated measurements of Band 7 matched between several gels.
The most intense bands are found in Gels 1 and 6 (Table 2) and in Figure 9 there is a very close correspondence of the rate of change between the top three bands of Gel 1 and Gel 6. It is likely that this effect is identifying a problem during autoradiography. At the higher intensifies the response of the film is not linear and is therefore reducing the slope of the curve. The higher rate of change between lane 1 and lane 2 of Gel 6 is almost identical to that between bands 3 and 4 of Gel 1 and Gel 4. This is simply because these bands are all of a similar intensity range which is below the point of the reduced response of the film. The alternative explanation which is that the range of the scanner was not sufficient is not considered to be the case because no discernible effect on the high intensity bands was noticed when scanning at different brightness settings (e.g. see Figures 4 and 6). Furthermore the slopes shown for the high density bands do not support theories regarding scanner range problems. The remainder of the bands tend to follow the same pattern of increase and when averaged out this approximates quite closely to a doubling series.
Using the Phoretix 1 D gel analysis software It was possible to adjust for nonlinearity of image capture by the Epson GT8000 scanner. It was also possible to identify problematic areas within the autoradiograms. Where the doubling series was not found the analysis clearly demonstrated this to be due to effects during hybridisation and autoradiography. Thus the conclusion that a standard commercial document scanner together with high quality gel analysis software can be used as a densitometer is supported. However, different coloured filters may be indicated and there may be a danger of damage to the scanner if wet gels are scanned unless the paten glass is sealed.
With regard to the software, it is important to note that the above analyses would not be as accurate, or as easy to do, with many other packages on the market. Phoretix uses a full 3D analysis (Length, Breadth and Depth), with the user having full and instant interaction with the image, the profile and a number of other supporting windows at all times. The presence of many automatic facilities, all with instant manual override, makes the task far easier and faster than is possible with many other packages. For example Gel 1 took no more than 5 minutes to fully analyse. All the data and calculations shown were done within Phoretix and then transported to Microsoft's Excel for presentation purposes. Furthermore, only a fraction of the software's capabilities have been used in the production of this article. The scanner used in this study produces 8 bit images. Both 12 and 16 bit images which are produced by several more sophisticated devices such as certain laser densitometers and radiation imaging devices can be analysed using Phoretix software.
Author: Bruce Venning (Phoretix International) 1996