- Email: [email protected]
Low-cost two-dimensional gel densitometry
ANALYTICAL
BIOCHEMISTRY
158,296301
Low-Cost
(1986)
Two-Dimensional
Gel Densitometry’
RICHARD M. LEVENSON, EDWARD V. MAYTIN,* E. Henry Keutmann Laboratories, Division of Endocrinology and Biophysics, University of Rochester School of Medicine
AND DONALD A. YOUNG~
and Metabolism, Departments of Medicine and Dentistry, Rochester, New York 14642
Received February 10, 1986 A major obstacle to full utilization of the powerful technique of two-dimensional (2-D) gel electrophoresis is the expense and complexity of quantifying the results. Using an analog-to-digital converter already present in the widely available Commodore 64 or Commodore 128 microcomputer, we have developed a 2-D gel densitometer (GELSCAN) which adds only $20.00 to the cost of the Commodore system (currently around $700.00). The system is designed to work with autoradiograms of 2-D gels. Spots of interest are identified visually and then positioned manually over a light source. A pinhole photoelectric sensor mounted in a hand-held, Plexiglas holder, or “mouse.” is briefly rubbed over each spot. Maximum density of the spot is determined and its value is converted to counts per minute via an internal calibration curve which corrects for the nonlinear response of film to radiation. Local spot backgrounds can be subtracted and values can be normalized between gels to adjust for variation in amount of radioactivity applied or in exposure time. Reproducibility is excellent and the technique has some practical as well as theoretical advantages over other more complicated approaches to 2-D gel densitometry. In addition, the GELSCAN system can also be used for scanning individual bands in 1-D gels. quantitation of “dot-blot” autoradiograms and other tasks involving transmission densitometry. Q 1986 Academic Press, Inc.
KEY WORDS: two-dimensional gel electrophoresis; isoelectric focusing: autoradiography; computer methods; densitometry; protein synthesis
The technique of two-dimensional gel electrophoresis pioneered by O’Farrell and others (1,2) represents a major methodological advance in protein separation, and is widely used throughout biological research. Current largescale applications of this method (“giant-gels”) followed by autoradiography permit the resolution of over 2000 proteins on a single autoradiogram (3) and over 3300 total cellular proteins have been demonstrated using overlapping pH ranges in the first dimension (4). Despite the obvious power of this technique, its use remains limited in part because of the ’ This work was supported by the James P. Wilmot Foundation and by NIH Grant AM 16 177. ‘Current address: Department of Medicine, Brown University, Roger Williams General Hospital, Providence, R.I. 02908. 3 To whom request for reprints should be addressed. 0003-2697/86 $3.00 Copyright 6 1986 by Academic Press. Inc. All rights of reproduction m any farm reserved.
expense and difficulty involved in analyzing and quantifying the results. Sophisticated computerized systems requiring powerful minicomputers or mainframes for digitizing, storing, and comparing entire autoradiograms have been designed (5); however, their cost places them out of reach of most laboratories’ budgets. Additionally, these systems limit the size of the gels that can be digitized, and their performance can be seriously degraded by problems in gel reproducibility. Less cumbersome systems may be more appropriate, since in many applications, only a few spots may vary and need quantitation (67). Even these simpler approaches commonly require expensive digitizing hardware and minicomputer computational capacity (7). Less expensive, video- and microcomputer-based systems typically have a restricted optical density range with relatively few (usually 64) gray-scale levels
294
LOW-COST
TWO-DIMENSIONAL
(8) and these still experience constraints on autoradiogram size. We have developed an extremely inexpensive and simple system (“GELSCAN”) which accomplishes the quantitation of spots previously identified by visual examination of the autoradiograms by using a novel arrangement of a pinhole photoelectric sensor mounted in a hand-held, movable Plexiglas holder (“mouse”). Signals from the photocell are connected to the analog-to-digital (A-D)4 converter already present in the inexpensive Commodore 64 or 128 computer. The hardware for the densitometer (excluding the slide projector pressed into service as a light source) costs less than $20.00. The complete system (Commodore computer plus peripherals) can be assembled for about $720.00 (U.S.). A system identical in concept can also be constructed using the Apple II-series of computers (9). METHODS 1. Two-Dimensional
Gel Electrophoresis
Confluent cultures of 3T3 cells were labeled with [35S]methionine, and the cells lysed in O’Farrell’s lysis buffer (1). Samples containing 80 kg of protein and 1,2,4, 8, and 15 million cpm were subjected to isoelectric focusing in tube gels (30 cm X 3.3 mm internal diameter) for 22 h at a final voltage of 2000 V. The firstdimension gels were extruded and annealed to the top of the second-dimension sodium dodecyl sulfate-polyacrylamide gels (30 X 30 X 0.075 cm) as previously described (3,4). After electrophoresis at 30 mA/gel for 12 h, the gels were fixed in 50% methanol, 12% acetic acid, dried, and exposed to Kodak X-AR film for 8 days. The developed autoradiograms were then scanned with the densitometer. An autoradiographic step tablet was prepared as described by Garrels (5) with the exception that the acrylamide strips were dried onto filter paper instead of being frozen for 4 Abbreviations used: A-D, analog-to-digital; I-D, onedimensional; 2-D, two-dimensional.
GEL DENSITOMETRY
295
later processing with flurographic enhancers. Briefly, 12.5% acrylamide was polymerized in layers containing 30 to 300,000 cpm/mm2 of dialyzed cellular protein labeled with [i4C]glycine; the strip was exposed to the film for 8 days to determine the radiation response of the film (Fig. 5, insert) and to establish the internal calibration curve for the densitometer. The calibration curve need only be set up once and can be used for autoradiograms exposed for various lengths of time. This is possible because the scanner does not integrate entire spots but rather determines relative labeling intensities for corresponding spots on multiple gels. For this purpose, only the film’s response to total radiation dose (which is independent of exposure time) is crucial (10). 2. Hardware
The requirements for the system are: a computer; light source; photodetector; circuit for connection of photodetector to computer; and peripheral devices such as a monitor, disk drive, and printer. The basic set-up is illustrated in Fig. 1. A. The computer. The principle of this densitometer can be realized using any computer with an analog-to-digital (A-D) interface. Such interfaces can be purchased or easily constructed. However, the Commodore 64 and 128 and the Apple II-series of computers have game-paddle ports connected to circuitry that can replace external A-D converters. In the case of the Commodore, these ports communicate with the 6581 Sound Interface Device @ID) chip which contains two 8-bit, AD potentiometer interfaces designed to interpret position information from paddle controls by converting changes in resistance into a digital signal (I 1). B. Electronics. A common, inexpensive ($10.00) NPN phototransistor (such as a Texas Instruments TIL99 or Sylvania ECG 3031) with sensitivity in the spectrum of light generated by incandescent bulbs is employed. It is mounted in a clear Plexiglas “mouse” with the walls of the l-mm pinhole internally
296
LEVENSON,
MAYTIN,
AND YOUNG
with the response of the A-D converter. The optical density range is only 0 to 1.2 OD when the voltage across the photoelectrode is limited to the 5 V supplied by the computer, but it is possible to extend the range to 2.2 optical density units by introducing two 9-V batteries into the circuit. However, when the 18 V are included, the lower range of optical densities becomes unreadable. Thus, a double-pole, double-throw switch is used to select low voltage for light spots and background readings, and high voltage for dark spots, permitting the entire range of densities to be scanned. Connections to another paddle lead, Pot X (Pin 9) allow the computer to detect the position of the switch, so that the raw numbers are conFIG. 1. Illustration of the GELSCAN densitometry sysverted to density measurements using the aptem. The individual components are identified as follows: propriate calibration curve. (We plan to re(A) hand-held “mouse” containing the photocell; (B) light source (in this case, a Leitz projecting microscope with a place the mechanical switch with a relay which mirror to direct the light vertically through the central automatically selects the correct voltage level area in the supporting glass plate; (C) resistors and batteries after a few milliseconds of scanning.) It may to match the photocell to the computer; (D) Commodore also be possible to extend the usable optical 64 computer; (E) monitor; (F) disk drive; (G) printer. A spot on a 2-D gel autoradiograph is positioned over the density range by adding a capacitor in parallel light source and is scanned by gently “rubbing” it with with the internal capacitor, as suggested by De the mouse for about I.5 s. Jong (9). By using the A-D converter in two different ranges, we divide our density signal into about 370 discrete steps, rather than the coated with opaque black paint to reduce in- 256 nominally available from an S-bit device. terference from extraneous light. The collector C. Light source. One of the more challengis connected to the computer’s 5-V power ing problems we faced was that of finding a supply (Pin 7) and the emitter is connected to bright, even, inexpensive incandescent light Pot Y (Pin 5) of the control port. The greater source. Illumination must be even over an area 1 cm2 since the spot to be the light impinging on the photoelectrode, the of approximately lower the resistance sensed by the A-D con- scanned (and its surrounding background verter. The resistance is digitized by a process area) is positioned by hand in the scanning based on the time constant of a capacitor in- region. Slide projectors have the appropriate light intensity, but many, such as the Kodak terposed between the Pot Y pin and ground. The capacitor is recharged by $5 V passing Carousel models, have filament configurations from the computer through the variable resis- and optics which give surprisingly uneven illumination. A Zeiss slide projector was very tance of the photoelectrode; the time required for recharge is measured and can be related to satisfactory but its owners did not agree to its light intensity. Bright light generates a signal indefinite loan; we are currently using a small of “0” and no light generates “255,” while in- Leitz projecting microscope without its lens. A ground-glass screen is inserted in the unit termediate levels of light result in generation near the condenser and the light is directed of the intervening numbers. vertically by means of a mirror up through Resistors arranged as shown in the circuit diagram (Fig. 2) were selected empirically to three layers of Scotch Magic tape or drafting mylar film which are positioned on the unmatch the properties of the photoelectrode
LOW-COST
PIN
TWO-DIMENSIONAL
9
66
kQ
297
GEL DENSITOMETRY
DPDT switch
3MQ
18V
Pot x
CONTROL PORT 1
; 4-3 530
6; :
I LOW
560
kQ
kQ
TIL99 PIN 5 0
i
Pot Y I
PHOTOCELL
FIG. 2. Circuit diagram for connecting the photocell to the Commodore 64 user port I. The pin assignments and their functions on the 658 1 SID chip are indicated. The values for the resistors were arrived at empirically and may have to be optimized for each computer. Pot Y receives the digital signal for spot density; Pot X is used to sense the position of the double-pole, double-throw (DPDT) switch so that the correct calibration curve is used. The photocell we use is silicon, NPN-type (Texas Instruments TIL99; equivalent is Sylvania ECG 3031) sensitive to portions of the visual and infrared spectrum emitted by incandescent bulbs; fluorescent lights are not suitable.
derside of the scanning aperture. This aperture is conveniently created by cutting a small circle in a sheet of undeveloped X-ray film which is taped to the glass plate supporting the autoradiogram. The film blocks enough light to protect the operator while permitting the pattern of spots on the autoradiogram to be seen. A constant voltage supply with potentiometer @taco E 10 1OVA) is used to adjust the intensity of the light, either to increase it for scanning overexposed gels, or to compensate for the dimming of the bulb as it nears the end of its working life. 3. Soft ware “GELSCAN,” the major program, is written in BASIC with two machine language subroutines for data acquisition and error checking. To perform a scan, the photoelectrodemouse is moved by hand over the spot for approximately 1.5 s while densities are detected and stored by the computer. The raw A-D output displays a rapid, periodic sinusoidal fluctuation of undetermined origin that is particularly noticeable with darker spots. Fortunately, this can be eliminated by aver-
aging the output values over several cycles. One machine language subroutine reads the potentiometer 256 times to obtain a smoothed average value, which is then stored in an array. This procedure is repeated 160 times for each spot. Occasionally, spurious “maximum density” signals (255’s) appear due to errors in reading the potentiometer value; these must be eliminated from the data or they will be scored as actual maxima. Thus, the 160 values are first checked for erroneous maxima with the second machine language subroutine which can quickly discard values that differ from their “downstream” neighbors by more than 1%. A true maximum is then identified and converted into either optical density or counts per minute by the appropriate calibration curve and interpolation subroutine. The local background for each spot is routinely subtracted from the scanned spot maximum: artifacts such as streaks and blurs can also be selectively subtracted at the operator’s discretion. Each spot is assigned a unique number and at least 900 spots can be scanned and kept in the computer’s memory with the current program. A virtually unlimited amount of spot data can be stored on magnetic disk.
298
LEVENSON,
MAYTIN,
The scanning results can be adjusted (normalized) to compensate for variations in the amount of radioactivity applied or in the exposure time. To accomplish this, a few “invariant” spots (i.e., spots which seem unaffected by the experimental conditions employed) are scanned and their densities are summed. For each autoradiogram, this sum serves as an indicator of overall density. To normalize a set of four autoradiograms (A, B, C, and D), the ratios of the “invariant” spot sum on gels B, C, and D to the invariant spot sum on gel A (i.e., ratios B/A, C/A, and D/A) are used to adjust the “experimental” spot densities on gels B, C, and D, normalizing them to the “reference” gel A. Another program, “GELHANDLER,” allows the averaging of results from several duplicate gels and calculates induction ratios (fold-increase or decrease versus control) for up to 12 autoradiograms. Results can also be exported into a spreadsheet program for further statistical analysis or graphical display.
AND
YOUNG
0 D
(reference)
FIG. 4. Correlation between nominal and scanned optical densities. A Kodak photographic step tablet was scanned on both the low and high scalesand the measured O.D. plotted against nominal O.D. Note that no discontinuity is seen when switching from the low to the high scale. The maximum optical density scanned here is 2.2 O.D. units; this can be increased by using a higher intensity light.
RESULTS
i/, 0
05
10
OPTICAL
15
20
25
DENSITY
FIG. 3. Digital signal generated in the Commodore 64’s A-D converter in response to scanning an optical density step tablet. Note the ranges of optical densities which can be scanned on either the low or the high voltage setting, the area of overlap between the two, and the nonlinear behavior exhibited by both. These raw numbers are automatically converted to optical density values by interpolation and calibration subroutines, as shown in Fig. 4.
The raw numbers digitally generated by the “GELSCAN” densitometer in response to different light intensities at both the low-voltage and high-voltage settings are shown in Fig. 3. This nonlinear response of the densitometer is linearized using a calibration curve and interpolation subroutine. Figure 4 shows the data generated by these subroutines when a Kodak optical density tablet is scanned; the smooth transition from the low-voltage to the highvoltage range is apparent. A scan involves the gentle rubbing of the “mouse” over the darkest area of the spot for about 1.5 s while data points are collected and the maximum density is determined by the computer. To determine reproducibility of this procedure, we scanned five spots of widely differing densities 10 times each. The coefficients of variation (standard deviation divided by the mean) of the maximum densities determined for each spot ranged from 1.7 to 5.4% (mean:
LOW-COST
TWO-DIMENSIONAL
299
GEL DENSITOMETRY
Materials and Methods. Twenty-two spots on each autoradiogram were scanned and the sums of their densities determined. Figure 6 illustrates the linear relationship between these summed values and the applied radioactivity on each gel and validates the use of maximum density as an estimate of total radioactivity present. (For further discussion of validity with individual spots, see below.) DISCUSSION
FIG. 5. Overall reproducibility of the giant 2-D gel system and densitometer. Two independent cultures of Swiss 3T3 cells were labeled with [35S]methionine, lysed. and the proteins separated on giant 2-D gels. Forty randomly selected spots on both autoradiograms were scanned and corresponding spot pairs were plotted. The correlation coefficient of the entire process from cell labeling to densitometry is r = 0.98. The insert shows the nonlinear response of the Kodak X-AR film to low and high exposures to radioactivity. This behavior is corrected for by the GELSCAN calibration curve.
3.4%, results not shown). The overall reproducibility of the system (from cell labeling through sample preparation, electrophoresis, autoradiography and quantitation) is demonstrated in Fig. 5. This figure is a correlation plot of maximum densities of corresponding spots from autoradiograms of two independent cultures of 3T3 cells. The insert demonstrates the nonlinear response of Kodak XAR film to the increasing doses of radioactivity on the autoradiography step-tablet used for establishing GELSCAN’s built-in calibration curve. Since GELSCAN determines the maximum density rather than the integrated density for each spot, it is important to demonstrate that maximum density can be linearly related to the total amount of radioactivity present. Autoradiograms of giant gels containing 1, 2, 4, 8, and 15 million cpm per gel of labeled cellular proteins were prepared as described in
We have found the “GELSCAN” densitometer described here to be invaluable in the interpretation of electrophoretic data. Its utility can be assessed in several recent publications (4,12,13). It is very inexpensive, adding less than $20.00 to the cost of a Commodore 64 computer or Apple II system, which can also be used for other tasks in the laboratory. Its simplicity and low cost is made possible by the arrangement of the light source, the use of a hand-held photocell, and the determination of maximum rather than integrated density.
500-
400
-j
300
2
0 L z z w
200
100
: t&L :
:
:
:
12
4
CPM
8
15
APPLIED
FIG. 6. Linear relationship between maximum spot density and overall spot radioactivity. Five giant gels were run with identical protein loads containing 1, 2, 4, 8, or I5 million cpm of labeled cellular proteins. Twenty-two spots on each gel were scanned, their densities summed, and these sums plotted against the applied radioactivity.
300
LEVENSON,
MAYTIN,
There are no constraints on gel size or shape, and no requirements for precise positioning of the autoradiograms or the use of laser scanners or video cameras. Although designed for 2-D gel densitometry, the scanner can also be used for scanning individual bands in 1-D gels, quantitation of dot-blot autoradiograms, and other tasks involving transmission densitometry. Unlike most other 2-D scanning systems, ours determines the maximum density rather than the integrated density of a spot. This is not useful if one is trying to compare the relative rates of synthesis of two different proteins which may differ in size and focusing characteristics. Instead of comparing synthetic rates for different proteins, we monitor the synthesis of corresponding proteins under different experimental conditions, and for this purpose, maximum density determination has some theoretical justification. The expression relating volume (i.e., integrated density) to shape for a well-resolved, Gaussian-distributed spot is V = xHab/e2 where a and b are the major and minor axes, and His the height, or maximum density (14). It can be seen that when spot size is constant, the integrated density is directly proportional to the maximum density for individual spots. (Spot size, it should be noted, refers not to the autoradiographic size which depends on labeling parameters, but the area of the gel actually occupied by the protein, demonstrable, for example, by silver stain.) The size of a spot on a 2-D gel is dependent not only on its individual focusing characteristics, but also on the amount of protein it contains. However, O’Farrell has shown that below a certain protein content threshold, changes in total protein content have only minor effects on spot size (1). Thus, to compare density maxima, we attempt to load roughly equivalent amounts of protein onto each gel, but in our 2-D system even fivefold differences in protein load seem
AND YOUNG
to have little effect on spot size for all except the most abundant proteins. Clearly, reliance on density maxima puts a premium on the running of good gels, since one cannot expect to compare the maximum density of a streaked protein with that of a well-focused one (although two identically streaked proteins can with some confidence be compared). Given reasonably good separations, however, the determination of maximum density does have some theoretical advantages over integrating spot densities: (1) maximum density can be determined without complicated algorithms necessary for resolving spot from background; (2) the use of density maxima may avoid systematic errors when integrating faint spots (10,15); (3) when neighboring spots or streaks encroach on the periphery of a spot, maximum density is less affected than integrated density, which is sensitive to overlapping throughout the area of a spot. The final answer to 2-D gel densitometry must await the arrival of far more computing power than is currently available to most laboratories. Until then, we believe that the “GELSCAN” system we describe here, which is not only inexpensive, rapid, simple and accurate, but is also well adapted to large, highresolution gels, will prove to be ofconsiderable utility. ACKNOWLEDGMENTS We wish to acknowledge the invaluable assistance of Dr. Jerry Miller in the development of the GELSCAN machine language subroutines, and Ms. Jan Dixon for drawing Fig. I and for careful review of the manuscript. REFERENCES 1. O’Farrell, P. H. ( 1975) .I Bid. Chem. 250,4007-402 I. Scheele, G. A. (1975)J. Biol. Chem. 250,5375-5385. 3. Young, D. A., Voris, B. P., Maytin, E. V., and Colbert, R. A. (1983) in Methods in Enzymology (Hirs, C. H. W., and Yimasheff, S. N., eds.), Vol. 91, Part 1. pp. 190-2 14. Academic Press, New York. 4. Levenson, R., Iwata, K., Klagsbrun, M., and Young, D. A. (1985) J. Biol. Chem. 260, 8056-8063. 2.
LOW-COST
TWO-DIMENSIONAL
5. Ganels, J. I. (1979) J. Biol. Chem 254, 7961-7977. 6. Voris, B. P., and Young, D. A. (I 98 I) J. Bio/. Chem. 256, 11,319-11,329. 7. Garrison, J. C., and Johnson, M. L. (1982) J. Biol. Chem. 257, 13,144-13,149. 8. Tracy, K. P., and Young, D. S. (1984) Clin. Chem. 30,462-465. 9. De Jong, M. L. (1985) Byte 10, 161-162. 10. Valiron, O., Lefkovits, I., Garderet, P., and Steinberg, C. (1984) Clin. Chem. 30, 1943-1946. Il. Commodore 64 Programmer’s Reference Guide
GEL DENSITOMETRY
12. 13. 14. 15.
301
(1982) Commodore Business Machines, pp. 469 and 472. Maytin, E. V., Colbert, R. A., and Young, D. A. (1985) J. Biol. Chem. 260,2384-2392. Colbert, R. A., and Young, D. A. (1986) Proc. Nut/. Acud. Sci. USA 83,72-76. Brown, W. T.. and Ezer, A. (1982) Clin. Chem. 28, 1041-1044. Bossinger. J., Miller, M. J., Vo, K.-P.. Geiduschek, E. P., and Xuong, N.-H. (1979) J. Biol. Chem. 254, 7986-7998.
BIOCHEMISTRY
158,296301
Low-Cost
(1986)
Two-Dimensional
Gel Densitometry’
RICHARD M. LEVENSON, EDWARD V. MAYTIN,* E. Henry Keutmann Laboratories, Division of Endocrinology and Biophysics, University of Rochester School of Medicine
AND DONALD A. YOUNG~
and Metabolism, Departments of Medicine and Dentistry, Rochester, New York 14642
Received February 10, 1986 A major obstacle to full utilization of the powerful technique of two-dimensional (2-D) gel electrophoresis is the expense and complexity of quantifying the results. Using an analog-to-digital converter already present in the widely available Commodore 64 or Commodore 128 microcomputer, we have developed a 2-D gel densitometer (GELSCAN) which adds only $20.00 to the cost of the Commodore system (currently around $700.00). The system is designed to work with autoradiograms of 2-D gels. Spots of interest are identified visually and then positioned manually over a light source. A pinhole photoelectric sensor mounted in a hand-held, Plexiglas holder, or “mouse.” is briefly rubbed over each spot. Maximum density of the spot is determined and its value is converted to counts per minute via an internal calibration curve which corrects for the nonlinear response of film to radiation. Local spot backgrounds can be subtracted and values can be normalized between gels to adjust for variation in amount of radioactivity applied or in exposure time. Reproducibility is excellent and the technique has some practical as well as theoretical advantages over other more complicated approaches to 2-D gel densitometry. In addition, the GELSCAN system can also be used for scanning individual bands in 1-D gels. quantitation of “dot-blot” autoradiograms and other tasks involving transmission densitometry. Q 1986 Academic Press, Inc.
KEY WORDS: two-dimensional gel electrophoresis; isoelectric focusing: autoradiography; computer methods; densitometry; protein synthesis
The technique of two-dimensional gel electrophoresis pioneered by O’Farrell and others (1,2) represents a major methodological advance in protein separation, and is widely used throughout biological research. Current largescale applications of this method (“giant-gels”) followed by autoradiography permit the resolution of over 2000 proteins on a single autoradiogram (3) and over 3300 total cellular proteins have been demonstrated using overlapping pH ranges in the first dimension (4). Despite the obvious power of this technique, its use remains limited in part because of the ’ This work was supported by the James P. Wilmot Foundation and by NIH Grant AM 16 177. ‘Current address: Department of Medicine, Brown University, Roger Williams General Hospital, Providence, R.I. 02908. 3 To whom request for reprints should be addressed. 0003-2697/86 $3.00 Copyright 6 1986 by Academic Press. Inc. All rights of reproduction m any farm reserved.
expense and difficulty involved in analyzing and quantifying the results. Sophisticated computerized systems requiring powerful minicomputers or mainframes for digitizing, storing, and comparing entire autoradiograms have been designed (5); however, their cost places them out of reach of most laboratories’ budgets. Additionally, these systems limit the size of the gels that can be digitized, and their performance can be seriously degraded by problems in gel reproducibility. Less cumbersome systems may be more appropriate, since in many applications, only a few spots may vary and need quantitation (67). Even these simpler approaches commonly require expensive digitizing hardware and minicomputer computational capacity (7). Less expensive, video- and microcomputer-based systems typically have a restricted optical density range with relatively few (usually 64) gray-scale levels
294
LOW-COST
TWO-DIMENSIONAL
(8) and these still experience constraints on autoradiogram size. We have developed an extremely inexpensive and simple system (“GELSCAN”) which accomplishes the quantitation of spots previously identified by visual examination of the autoradiograms by using a novel arrangement of a pinhole photoelectric sensor mounted in a hand-held, movable Plexiglas holder (“mouse”). Signals from the photocell are connected to the analog-to-digital (A-D)4 converter already present in the inexpensive Commodore 64 or 128 computer. The hardware for the densitometer (excluding the slide projector pressed into service as a light source) costs less than $20.00. The complete system (Commodore computer plus peripherals) can be assembled for about $720.00 (U.S.). A system identical in concept can also be constructed using the Apple II-series of computers (9). METHODS 1. Two-Dimensional
Gel Electrophoresis
Confluent cultures of 3T3 cells were labeled with [35S]methionine, and the cells lysed in O’Farrell’s lysis buffer (1). Samples containing 80 kg of protein and 1,2,4, 8, and 15 million cpm were subjected to isoelectric focusing in tube gels (30 cm X 3.3 mm internal diameter) for 22 h at a final voltage of 2000 V. The firstdimension gels were extruded and annealed to the top of the second-dimension sodium dodecyl sulfate-polyacrylamide gels (30 X 30 X 0.075 cm) as previously described (3,4). After electrophoresis at 30 mA/gel for 12 h, the gels were fixed in 50% methanol, 12% acetic acid, dried, and exposed to Kodak X-AR film for 8 days. The developed autoradiograms were then scanned with the densitometer. An autoradiographic step tablet was prepared as described by Garrels (5) with the exception that the acrylamide strips were dried onto filter paper instead of being frozen for 4 Abbreviations used: A-D, analog-to-digital; I-D, onedimensional; 2-D, two-dimensional.
GEL DENSITOMETRY
295
later processing with flurographic enhancers. Briefly, 12.5% acrylamide was polymerized in layers containing 30 to 300,000 cpm/mm2 of dialyzed cellular protein labeled with [i4C]glycine; the strip was exposed to the film for 8 days to determine the radiation response of the film (Fig. 5, insert) and to establish the internal calibration curve for the densitometer. The calibration curve need only be set up once and can be used for autoradiograms exposed for various lengths of time. This is possible because the scanner does not integrate entire spots but rather determines relative labeling intensities for corresponding spots on multiple gels. For this purpose, only the film’s response to total radiation dose (which is independent of exposure time) is crucial (10). 2. Hardware
The requirements for the system are: a computer; light source; photodetector; circuit for connection of photodetector to computer; and peripheral devices such as a monitor, disk drive, and printer. The basic set-up is illustrated in Fig. 1. A. The computer. The principle of this densitometer can be realized using any computer with an analog-to-digital (A-D) interface. Such interfaces can be purchased or easily constructed. However, the Commodore 64 and 128 and the Apple II-series of computers have game-paddle ports connected to circuitry that can replace external A-D converters. In the case of the Commodore, these ports communicate with the 6581 Sound Interface Device @ID) chip which contains two 8-bit, AD potentiometer interfaces designed to interpret position information from paddle controls by converting changes in resistance into a digital signal (I 1). B. Electronics. A common, inexpensive ($10.00) NPN phototransistor (such as a Texas Instruments TIL99 or Sylvania ECG 3031) with sensitivity in the spectrum of light generated by incandescent bulbs is employed. It is mounted in a clear Plexiglas “mouse” with the walls of the l-mm pinhole internally
296
LEVENSON,
MAYTIN,
AND YOUNG
with the response of the A-D converter. The optical density range is only 0 to 1.2 OD when the voltage across the photoelectrode is limited to the 5 V supplied by the computer, but it is possible to extend the range to 2.2 optical density units by introducing two 9-V batteries into the circuit. However, when the 18 V are included, the lower range of optical densities becomes unreadable. Thus, a double-pole, double-throw switch is used to select low voltage for light spots and background readings, and high voltage for dark spots, permitting the entire range of densities to be scanned. Connections to another paddle lead, Pot X (Pin 9) allow the computer to detect the position of the switch, so that the raw numbers are conFIG. 1. Illustration of the GELSCAN densitometry sysverted to density measurements using the aptem. The individual components are identified as follows: propriate calibration curve. (We plan to re(A) hand-held “mouse” containing the photocell; (B) light source (in this case, a Leitz projecting microscope with a place the mechanical switch with a relay which mirror to direct the light vertically through the central automatically selects the correct voltage level area in the supporting glass plate; (C) resistors and batteries after a few milliseconds of scanning.) It may to match the photocell to the computer; (D) Commodore also be possible to extend the usable optical 64 computer; (E) monitor; (F) disk drive; (G) printer. A spot on a 2-D gel autoradiograph is positioned over the density range by adding a capacitor in parallel light source and is scanned by gently “rubbing” it with with the internal capacitor, as suggested by De the mouse for about I.5 s. Jong (9). By using the A-D converter in two different ranges, we divide our density signal into about 370 discrete steps, rather than the coated with opaque black paint to reduce in- 256 nominally available from an S-bit device. terference from extraneous light. The collector C. Light source. One of the more challengis connected to the computer’s 5-V power ing problems we faced was that of finding a supply (Pin 7) and the emitter is connected to bright, even, inexpensive incandescent light Pot Y (Pin 5) of the control port. The greater source. Illumination must be even over an area 1 cm2 since the spot to be the light impinging on the photoelectrode, the of approximately lower the resistance sensed by the A-D con- scanned (and its surrounding background verter. The resistance is digitized by a process area) is positioned by hand in the scanning based on the time constant of a capacitor in- region. Slide projectors have the appropriate light intensity, but many, such as the Kodak terposed between the Pot Y pin and ground. The capacitor is recharged by $5 V passing Carousel models, have filament configurations from the computer through the variable resis- and optics which give surprisingly uneven illumination. A Zeiss slide projector was very tance of the photoelectrode; the time required for recharge is measured and can be related to satisfactory but its owners did not agree to its light intensity. Bright light generates a signal indefinite loan; we are currently using a small of “0” and no light generates “255,” while in- Leitz projecting microscope without its lens. A ground-glass screen is inserted in the unit termediate levels of light result in generation near the condenser and the light is directed of the intervening numbers. vertically by means of a mirror up through Resistors arranged as shown in the circuit diagram (Fig. 2) were selected empirically to three layers of Scotch Magic tape or drafting mylar film which are positioned on the unmatch the properties of the photoelectrode
LOW-COST
PIN
TWO-DIMENSIONAL
9
66
kQ
297
GEL DENSITOMETRY
DPDT switch
3MQ
18V
Pot x
CONTROL PORT 1
; 4-3 530
6; :
I LOW
560
kQ
kQ
TIL99 PIN 5 0
i
Pot Y I
PHOTOCELL
FIG. 2. Circuit diagram for connecting the photocell to the Commodore 64 user port I. The pin assignments and their functions on the 658 1 SID chip are indicated. The values for the resistors were arrived at empirically and may have to be optimized for each computer. Pot Y receives the digital signal for spot density; Pot X is used to sense the position of the double-pole, double-throw (DPDT) switch so that the correct calibration curve is used. The photocell we use is silicon, NPN-type (Texas Instruments TIL99; equivalent is Sylvania ECG 3031) sensitive to portions of the visual and infrared spectrum emitted by incandescent bulbs; fluorescent lights are not suitable.
derside of the scanning aperture. This aperture is conveniently created by cutting a small circle in a sheet of undeveloped X-ray film which is taped to the glass plate supporting the autoradiogram. The film blocks enough light to protect the operator while permitting the pattern of spots on the autoradiogram to be seen. A constant voltage supply with potentiometer @taco E 10 1OVA) is used to adjust the intensity of the light, either to increase it for scanning overexposed gels, or to compensate for the dimming of the bulb as it nears the end of its working life. 3. Soft ware “GELSCAN,” the major program, is written in BASIC with two machine language subroutines for data acquisition and error checking. To perform a scan, the photoelectrodemouse is moved by hand over the spot for approximately 1.5 s while densities are detected and stored by the computer. The raw A-D output displays a rapid, periodic sinusoidal fluctuation of undetermined origin that is particularly noticeable with darker spots. Fortunately, this can be eliminated by aver-
aging the output values over several cycles. One machine language subroutine reads the potentiometer 256 times to obtain a smoothed average value, which is then stored in an array. This procedure is repeated 160 times for each spot. Occasionally, spurious “maximum density” signals (255’s) appear due to errors in reading the potentiometer value; these must be eliminated from the data or they will be scored as actual maxima. Thus, the 160 values are first checked for erroneous maxima with the second machine language subroutine which can quickly discard values that differ from their “downstream” neighbors by more than 1%. A true maximum is then identified and converted into either optical density or counts per minute by the appropriate calibration curve and interpolation subroutine. The local background for each spot is routinely subtracted from the scanned spot maximum: artifacts such as streaks and blurs can also be selectively subtracted at the operator’s discretion. Each spot is assigned a unique number and at least 900 spots can be scanned and kept in the computer’s memory with the current program. A virtually unlimited amount of spot data can be stored on magnetic disk.
298
LEVENSON,
MAYTIN,
The scanning results can be adjusted (normalized) to compensate for variations in the amount of radioactivity applied or in the exposure time. To accomplish this, a few “invariant” spots (i.e., spots which seem unaffected by the experimental conditions employed) are scanned and their densities are summed. For each autoradiogram, this sum serves as an indicator of overall density. To normalize a set of four autoradiograms (A, B, C, and D), the ratios of the “invariant” spot sum on gels B, C, and D to the invariant spot sum on gel A (i.e., ratios B/A, C/A, and D/A) are used to adjust the “experimental” spot densities on gels B, C, and D, normalizing them to the “reference” gel A. Another program, “GELHANDLER,” allows the averaging of results from several duplicate gels and calculates induction ratios (fold-increase or decrease versus control) for up to 12 autoradiograms. Results can also be exported into a spreadsheet program for further statistical analysis or graphical display.
AND
YOUNG
0 D
(reference)
FIG. 4. Correlation between nominal and scanned optical densities. A Kodak photographic step tablet was scanned on both the low and high scalesand the measured O.D. plotted against nominal O.D. Note that no discontinuity is seen when switching from the low to the high scale. The maximum optical density scanned here is 2.2 O.D. units; this can be increased by using a higher intensity light.
RESULTS
i/, 0
05
10
OPTICAL
15
20
25
DENSITY
FIG. 3. Digital signal generated in the Commodore 64’s A-D converter in response to scanning an optical density step tablet. Note the ranges of optical densities which can be scanned on either the low or the high voltage setting, the area of overlap between the two, and the nonlinear behavior exhibited by both. These raw numbers are automatically converted to optical density values by interpolation and calibration subroutines, as shown in Fig. 4.
The raw numbers digitally generated by the “GELSCAN” densitometer in response to different light intensities at both the low-voltage and high-voltage settings are shown in Fig. 3. This nonlinear response of the densitometer is linearized using a calibration curve and interpolation subroutine. Figure 4 shows the data generated by these subroutines when a Kodak optical density tablet is scanned; the smooth transition from the low-voltage to the highvoltage range is apparent. A scan involves the gentle rubbing of the “mouse” over the darkest area of the spot for about 1.5 s while data points are collected and the maximum density is determined by the computer. To determine reproducibility of this procedure, we scanned five spots of widely differing densities 10 times each. The coefficients of variation (standard deviation divided by the mean) of the maximum densities determined for each spot ranged from 1.7 to 5.4% (mean:
LOW-COST
TWO-DIMENSIONAL
299
GEL DENSITOMETRY
Materials and Methods. Twenty-two spots on each autoradiogram were scanned and the sums of their densities determined. Figure 6 illustrates the linear relationship between these summed values and the applied radioactivity on each gel and validates the use of maximum density as an estimate of total radioactivity present. (For further discussion of validity with individual spots, see below.) DISCUSSION
FIG. 5. Overall reproducibility of the giant 2-D gel system and densitometer. Two independent cultures of Swiss 3T3 cells were labeled with [35S]methionine, lysed. and the proteins separated on giant 2-D gels. Forty randomly selected spots on both autoradiograms were scanned and corresponding spot pairs were plotted. The correlation coefficient of the entire process from cell labeling to densitometry is r = 0.98. The insert shows the nonlinear response of the Kodak X-AR film to low and high exposures to radioactivity. This behavior is corrected for by the GELSCAN calibration curve.
3.4%, results not shown). The overall reproducibility of the system (from cell labeling through sample preparation, electrophoresis, autoradiography and quantitation) is demonstrated in Fig. 5. This figure is a correlation plot of maximum densities of corresponding spots from autoradiograms of two independent cultures of 3T3 cells. The insert demonstrates the nonlinear response of Kodak XAR film to the increasing doses of radioactivity on the autoradiography step-tablet used for establishing GELSCAN’s built-in calibration curve. Since GELSCAN determines the maximum density rather than the integrated density for each spot, it is important to demonstrate that maximum density can be linearly related to the total amount of radioactivity present. Autoradiograms of giant gels containing 1, 2, 4, 8, and 15 million cpm per gel of labeled cellular proteins were prepared as described in
We have found the “GELSCAN” densitometer described here to be invaluable in the interpretation of electrophoretic data. Its utility can be assessed in several recent publications (4,12,13). It is very inexpensive, adding less than $20.00 to the cost of a Commodore 64 computer or Apple II system, which can also be used for other tasks in the laboratory. Its simplicity and low cost is made possible by the arrangement of the light source, the use of a hand-held photocell, and the determination of maximum rather than integrated density.
500-
400
-j
300
2
0 L z z w
200
100
: t&L :
:
:
:
12
4
CPM
8
15
APPLIED
FIG. 6. Linear relationship between maximum spot density and overall spot radioactivity. Five giant gels were run with identical protein loads containing 1, 2, 4, 8, or I5 million cpm of labeled cellular proteins. Twenty-two spots on each gel were scanned, their densities summed, and these sums plotted against the applied radioactivity.
300
LEVENSON,
MAYTIN,
There are no constraints on gel size or shape, and no requirements for precise positioning of the autoradiograms or the use of laser scanners or video cameras. Although designed for 2-D gel densitometry, the scanner can also be used for scanning individual bands in 1-D gels, quantitation of dot-blot autoradiograms, and other tasks involving transmission densitometry. Unlike most other 2-D scanning systems, ours determines the maximum density rather than the integrated density of a spot. This is not useful if one is trying to compare the relative rates of synthesis of two different proteins which may differ in size and focusing characteristics. Instead of comparing synthetic rates for different proteins, we monitor the synthesis of corresponding proteins under different experimental conditions, and for this purpose, maximum density determination has some theoretical justification. The expression relating volume (i.e., integrated density) to shape for a well-resolved, Gaussian-distributed spot is V = xHab/e2 where a and b are the major and minor axes, and His the height, or maximum density (14). It can be seen that when spot size is constant, the integrated density is directly proportional to the maximum density for individual spots. (Spot size, it should be noted, refers not to the autoradiographic size which depends on labeling parameters, but the area of the gel actually occupied by the protein, demonstrable, for example, by silver stain.) The size of a spot on a 2-D gel is dependent not only on its individual focusing characteristics, but also on the amount of protein it contains. However, O’Farrell has shown that below a certain protein content threshold, changes in total protein content have only minor effects on spot size (1). Thus, to compare density maxima, we attempt to load roughly equivalent amounts of protein onto each gel, but in our 2-D system even fivefold differences in protein load seem
AND YOUNG
to have little effect on spot size for all except the most abundant proteins. Clearly, reliance on density maxima puts a premium on the running of good gels, since one cannot expect to compare the maximum density of a streaked protein with that of a well-focused one (although two identically streaked proteins can with some confidence be compared). Given reasonably good separations, however, the determination of maximum density does have some theoretical advantages over integrating spot densities: (1) maximum density can be determined without complicated algorithms necessary for resolving spot from background; (2) the use of density maxima may avoid systematic errors when integrating faint spots (10,15); (3) when neighboring spots or streaks encroach on the periphery of a spot, maximum density is less affected than integrated density, which is sensitive to overlapping throughout the area of a spot. The final answer to 2-D gel densitometry must await the arrival of far more computing power than is currently available to most laboratories. Until then, we believe that the “GELSCAN” system we describe here, which is not only inexpensive, rapid, simple and accurate, but is also well adapted to large, highresolution gels, will prove to be ofconsiderable utility. ACKNOWLEDGMENTS We wish to acknowledge the invaluable assistance of Dr. Jerry Miller in the development of the GELSCAN machine language subroutines, and Ms. Jan Dixon for drawing Fig. I and for careful review of the manuscript. REFERENCES 1. O’Farrell, P. H. ( 1975) .I Bid. Chem. 250,4007-402 I. Scheele, G. A. (1975)J. Biol. Chem. 250,5375-5385. 3. Young, D. A., Voris, B. P., Maytin, E. V., and Colbert, R. A. (1983) in Methods in Enzymology (Hirs, C. H. W., and Yimasheff, S. N., eds.), Vol. 91, Part 1. pp. 190-2 14. Academic Press, New York. 4. Levenson, R., Iwata, K., Klagsbrun, M., and Young, D. A. (1985) J. Biol. Chem. 260, 8056-8063. 2.
LOW-COST
TWO-DIMENSIONAL
5. Ganels, J. I. (1979) J. Biol. Chem 254, 7961-7977. 6. Voris, B. P., and Young, D. A. (I 98 I) J. Bio/. Chem. 256, 11,319-11,329. 7. Garrison, J. C., and Johnson, M. L. (1982) J. Biol. Chem. 257, 13,144-13,149. 8. Tracy, K. P., and Young, D. S. (1984) Clin. Chem. 30,462-465. 9. De Jong, M. L. (1985) Byte 10, 161-162. 10. Valiron, O., Lefkovits, I., Garderet, P., and Steinberg, C. (1984) Clin. Chem. 30, 1943-1946. Il. Commodore 64 Programmer’s Reference Guide
GEL DENSITOMETRY
12. 13. 14. 15.
301
(1982) Commodore Business Machines, pp. 469 and 472. Maytin, E. V., Colbert, R. A., and Young, D. A. (1985) J. Biol. Chem. 260,2384-2392. Colbert, R. A., and Young, D. A. (1986) Proc. Nut/. Acud. Sci. USA 83,72-76. Brown, W. T.. and Ezer, A. (1982) Clin. Chem. 28, 1041-1044. Bossinger. J., Miller, M. J., Vo, K.-P.. Geiduschek, E. P., and Xuong, N.-H. (1979) J. Biol. Chem. 254, 7986-7998.