|Home | About | Journals | Submit | Contact Us | Français|
Digital image analysis of the separation results of colorless analytes on thin-layer chromatography (TLC) plates usually involves using specially tailored software to analyze the images generated from either a UV scanner or UV lamp station with a digital camera or a densitometer. Here, a low-cost alternative setup for quantitative TLC-digital image analysis is demonstrated using a universal staining reagent (iodine vapor), an office scanner and a commonly available software (Microsoft Paint) for analysis of red, green and blue colors (RGB values). Urinary creatinine is used as a model analyte to represent a sample in complicated biological matrices. Separation was carried out on a silica gel plate using a butanol–NH4OH–H2O (40 : 10 : 50, v/v) mobile phase with a 6-cm solvent front. It is important that the TLC plate be stained evenly and with sufficient staining time. Staining the TLC plate in a 23.4 × 18.8 × 6.8 cm chamber containing about 70 g iodine crystals yielded comparable results for the staining times of 30–60 min. The Green value offered the best results in the linear working range (0.0810–0.9260 mg/mL) and precision (2.03% RSD, n = 10). The detection limit was found to be 0.24 µg per 3 µL spot. Urinary creatinine concentrations determined by TLC-digital image analysis using the green value calibration graph agree well with results obtained from high-pressure liquid chromatography (HPLC).
Thin-layer chromatography (TLC) is considered a sustainable analytical technique, and is the method of choice in many laboratories with a limited budget. Its simplistic setup and low cost with no maintenance requirements are the main advantages over other formats of chromatographic techniques such as high-pressure liquid chromatography (HPLC) and gas chromatography (GC) (1). In the past, TLC was used only for qualitative and semi-quantitative analyses. The recent advancements in digital image technologies enable a more detailed quantitative analysis from the image of the TLC plate. There have been many reports on quantitative determination based on TLC-image analysis using a flatbed scanner to record the image of the colored analyte spots. For colorless spots, a special UV scanner is available (2), although most works reported the use of a densitometer or a digital camera together with a UV lamp station to record the intensity of the grayish spots. Then, either a tailor-made or commercial software [such as TLSee (Alfatech Spa, Italy), Sorbfil (Sorbpolymer, Russia) or IGOR (WaveMetrics, USA)] was used to convert the intensity of each spot into a peak profile or 3D image with height and area that correlate to the concentration of the analyte (3–9). Only a few groups (10, 11) reported the use of staining reagents to make a colorless analyte spot visible that could be recorded by a flatbed scanner. However, the tailor-made software or special data analysis software [JustTLC (Sweday, Sweden), IGOR] were still required for the image analysis step. These types of commercial software and UV station setups are not commonly available in many laboratories.
In this work, we investigate the possibility of using only easy to find materials/reagents and software to perform TLC-image analysis. Urinary creatinine is selected as a model analyte. Creatinine is colorless and is secreted in urine at an easily detectable level. It is a common biomarker of renal function and has been used as an indicator of urine tampering or dilution in routine drug tests as well as an internal standard for analysis of other substances such as by the analyte–creatinine ratio (12, 13). In addition, urine collection is noninvasive and urine exhibits complex matrices which will help demonstrate the performance of the proposed TLC-image analysis method. The proposed method used commonly available I2 vapor as a staining reagent and the software Microsoft Paint (which is included with all versions of Microsoft Windows) to analyze the images of the TLC plate that were recorded with an office scanner. The stained spots were brownish in color with intensity depending on concentration. Based on the fact that the primary colors of light are red, green and blue, the intensity of each spot can be revealed as red, green and blue values (RGB). Various parameters that may cause error were investigated in detail. These include the sample preparation method, staining chamber geometry, staining time and the reading of RGB values and evaluation of their usability. After optimization, urine samples were analyzed for their creatinine contents using both the proposed TLC method and HPLC for comparison. Benefits from the method include (1) extending application of TLC to quantitative analysis while maintaining TLC operation at a low cost using easily available reagents, equipment and software, (2) flexibility of performing data analysis at a later time without the need to do data analysis before the staining color fades away and (3) enabling record keeping of data for future reference.
Polyester-backed silica gel TLC plates (5 × 20 and 20 × 20 cm) with a fluorescence indicator (UV254) were purchased from Sorbitech Sorbent Technology. Creatinine and uric acid (Sigma-Aldrich) were prepared separately. Uric acid solution was prepared in basic solution (water with NaOH added to obtain a pH of ~10), and creatinine was prepared in deionized (DI) water. Various mobile phases for TLC were prepared from different volume ratios of butanol, 1-propanol, 2- propanol (Fisher), ammonium hydroxide, glacial acetic acid (ACS reagent, Sigma-Aldrich) and DI water. A micropipette (size 0.5–5 μL, Fisher) was used to introduce the standard/sample spot (3 µL) onto the TLC plate. The developing chamber was a 27.0 × 26.5 × 7.0 cm rectangular tank (Aldrich). The most suitable staining chamber was a 23.4 × 18.8 × 6.8 cm glass storage container with a plastic lid (Pyrex Simply Store), in which its bottom surface was covered with iodine crystals (Aldrich). The stained TLC plate was scanned with a scanner (Cannon MG2520). Spots on the image were analyzed using Microsoft Paint and Microsoft Power Point programs.
The HPLC system used for comparison of results was the Ultimate 3000 (Dionex), equipped with a C18 packed column with 120 Å particle size and 4.6 mm i.d. × 100 mm length (Acclaim). The mobile phase for the isocratic separation of creatinine was a 0.05 M ammonium phosphate buffer (pH 7.4) (14). It was prepared by dissolving 2.2195 g (NH4)H2PO4 and 4.0524 g (NH4)2HPO4 in ~800 mL DI water. Its pH was adjusted by adding 1 M NaOH, and the volume of the buffer solution was made up to 1.00 L with DI water.
Upon obtaining permission from the Institutional Review Board (IRB) and consent from the volunteers, urine samples were collected from each of four volunteers in glass vials on multiple days, at ~10 mL each time. Some of the samples from different volunteers were mixed at known volume ratio or spiked with a small amount of standard creatinine to create a larger number of samples for different studies. These samples were filtered with 0.22 µm mixed cellulose ester (MCE) membrane syringe filters (25 mm, Fisher) and diluted with DI water to the desired dilution factor prior to the analysis. All samples were analyzed within 24 h of collection.
The TLC plate was cut into the desired size. Standards and samples were spotted 1.2 cm above the end of the plate at a spacing of 1 cm. Each solution was spotted only once to reduce the imprecision of repetitive spotting. A 3 µL spot volume was determined to be the most suitable amount in order to obtain small diameter spots while having a sufficient amount of analyte to be seen in the digital image. At this condition, a TLC plate of 20 cm width could accommodate 19 standard and sample spots. After drying the solvent with a hair dryer, the TLC plate was placed into the developing chamber containing 40 mL of the mobile phase (~0.7 cm from the bottom of the chamber). The plate was allowed to develop up to the desired solvent front. Then, the plate was removed from the chamber and blown dry with the hair dryer in the hood prior to putting it in the staining chamber. After staining for a desired period of time, the TLC plate was scanned at 300 dpi and its image was recorded in the personal computer. It was reported that the resolution setting of the scanner (72–1,200 dpi) does not significantly affect the relative intensities of the spots on the image (10); therefore, the lowest resolution setting of the scanner (300 dpi) was used for rapid scanning. To prevent I2 vapor from staining the surface of the scanner, the stained TLC plate was sandwiched in between two transparent plastic sheets during scanning. The RGB values of each spot were analyzed using Microsoft Paint program.
In several published HPLC separations of urine samples, creatinine has been studied with uric acid. In some separation systems, the two peaks are distinguishable but very close to each other (14, 15). To make sure that uric acid will not interfere with creatinine analysis on TLC, various mobile phases (as shown in Figure Figure1)1) were tested for separation of urinary creatinine by comparing the retention factor (Rf value) using a standard solution of creatinine and uric acid. In this preliminary experiment, 5 × 12 and 5 × 8 cm TLC plates were used to accommodate four spots: creatinine standard, uric acid standard, 1 : 1v/v creatinine:uric acid mixture and urine sample. A round glass staining chamber was used in this study. Separation distances (from starting spot to solvent front) of 6 and 10 cm were tested and it was found that the distance of the solvent front did not significantly affect the distance traveled by uric acid and creatinine. Therefore, to shorten analysis time, 6 cm was chosen for further studies.
On the other hand, mobile phase composition has a pronounced effect on the distance traveled by the analytes. As presented in Figure Figure1,1, the separated spots had oval shape when using 1- or 2- propanol in the mobile phase. Propanol also causes the creatinine spot to travel much faster and closer to the solvent front. It is commonly known that a small amount of acetic acid (CH3COOH) or ammonium hydroxide (NH4OH) may help adjust the shape and reduce the concaveness of the separated spots. However, adding CH3COOH in propanol caused the uric acid to have band tailing from the starting point to the solvent front (Figure (Figure1A1A and B). This can cause error in the analysis of creatinine in a urine sample if it contains a high concentration of uric acid. Addition of NH4OH in propanol caused a darker and uneven background (Figure (Figure11C–E).
Separated spots had a more rounded shape when using butanol. Spots also traveled much slower, with uric acid almost unmoved from the original spot. The overall background is also cleaner when using butanol. Addition of a small amount of NH4OH helped to make the separated spots have a more well-defined border when compared with addition of glacial CH3COOH. However, addition of NH4OH in butanol (Figure (Figure1H–J)1H–J) shows a white spot in the middle of uric acid and creatinine standards, but not in the mixture and urine sample. Later experiments revealed that due to insufficient staining time, high concentration individual uric acid and creatinine standards were not completely stained. Uric acid and creatinine were more diluted in the mixture and urine sample, and therefore, were stained more completely without the white spot in the middle. The optimization of working concentration range and staining time could eliminate this problem. Butanol–NH4OH–H2O at a 40 : 10 : 50 v/v ratio was chosen for further experiments because it offered a cleaner TLC plate background after staining with I2 and required a lower volume of organic solvents used in the mobile phase composition, with the additional benefit of slightly lower cost NH4OH when compared with glacial CH3COOH (16). With this mobile phase, the separation was completed in 1 h 25 min. Lines at the solvent front that appeared in all the TLC plates most likely come from additives or impurities on the TLC plate which can be removed by developing an unused TLC plate in the mobile phase and drying it prior to use for analysis (results not shown). However, to save time, this step was not carried out because the creatinine spot of interest was far from these lines at the solvent front and the analysis of creatinine was not affected.
Staining is an important step to reveal the analyte spots for the subsequent image analysis. To keep the main objective of this study on using easily available resources, only two common staining reagents, iodine (I2) vapor and potassium permanganate (KMnO4), were considered. An additional heating step is necessary after submerging the TLC plate in the KMnO4 solution. This normally requires a heat gun, which is not always available in every laboratory. Therefore, three alternative methods were tested: (1) heating with a hair dryer, (2) heating in the oven at 100°C for 10 min and (3) heating on the hot plate. The first two heating methods did not show any changes. The third method turned the TLC plate background an uneven purplish color and spots of the analytes could not be observed. With the unsuccessful additional heating step, KMnO4 staining was discontinued.
Staining with I2 vapor is a simpler one-step process, but the selection of staining chambers as well as ensuring that the TLC plate is evenly exposed to the I2 vapor is critical. Uneven staining can cause imprecision in the analysis (11). When placing the TLC plate vertically in the round staining chamber saturated with I2 vapor, the plate was not evenly stained. The bottom corners of the plate, closer to iodine crystals were darker than the middle area; see Figure Figure2A.2A. A shallow rectangular container with the bottom surface covered with I2 crystals (70.0 g) was tested. Pieces of two-faced tape were adhered onto the polyester back of the TLC plate. The plate was then secured onto the inside of the plastic lid of the I2 chamber. When putting the lid on the chamber, the TLC plate was faced down and exposed to the I2 vapor. This way, the whole TLC plate was parallel and was positioned at the same distance from the I2 crystals that covered the bottom of the chamber and was evenly stained; see Figure Figure2B.2B. The amount of I2 in the chamber was checked daily by weighing the staining chamber containing I2 and subtracting the initial weight of the empty chamber. The weight of I2 was kept within 70.0 ± 2.0 g, to ensure constant vapor saturation comparable to previous experiments.
The longer the staining time, the darker the TLC plate became, thus making the low-concentration spots more visible, as shown in Figure Figure2C.2C. To study the effect of the staining time on the working range of creatinine in more detail, the TLC plate was removed from the staining chamber every 5 min and was scanned at 300 dpi. Then, the plate was placed back in the staining chamber, and the process was repeated up to 60 min. The RGB values were plotted with the concentrations of creatinine standard (0.0810–1.1575 mg/mL), as shown in Figure Figure3A–C.3A–C. The red graphs could not quite distinguish low concentrations of creatinine. In contrast, the green and blue values could distinguish low concentrations, but they leveled off at high concentrations. At too short of a staining time (i.e., 5 min), not all the spots are completely stained and low concentrations could not be clearly observed. Increasing staining time increases the intensity of the spot and, as a result, increases the slope of the graphs. Overlaying the calibration graphs obtained from various staining times reveals that slopes of the graphs remain fairly constant at longer staining times, which indicates more complete staining. Table TableII presents linear regression equations obtained from RGB values at various staining times for comparison. The slopes of blue value fluctuated more than those of other colors. In addition, for clarity, Figure Figure3D3D also shows the comparison of the red, green and blue calibration graphs obtained from a staining time of 50 min. The linear range of blue is the narrowest and would not be very useful. Slopes of red and green values are rather constant at the staining time of 30 min and longer (see slope values in Table TableI).I). Although red values cover a wider linear range when compared with green values (0.2315–1.1575 vs 0.0810–0.9260 mg/mL; see Figure Figure3D),3D), the green calibration graph was proved to be more accurate when using it to determine urinary creatinine concentration when compared with the results obtained from HPLC. This is most likely because green values cover the low-concentration region (0.0810–0.2315 mg/mL), whereas red values do not. Therefore, the green calibration graph was selected for further experiments.
To extend the life time of the HPLC column, filtration and dilution of samples are common steps to remove particulates and some proteins from the samples prior to analyzing them. For TLC, even though there is no such worry of column clogging as in HPLC, particulates and proteins in untreated urine samples may cause low separation resolution and error in RGB values of the creatinine spot. To find the most suitable sample preparation, the following methods were tested. Method 1, the urine sample was diluted to the desired dilution factor with DI water and analyzed without further treatment. Method 2, the urine sample was filtered with a syringe filter (0.22 µm MCE membrane, Fisher) and then diluted with DI water prior to the analysis. Method 3, 10 µL of phosphoric acid was added to a 2 mL urine sample to precipitate proteins which were removed by centrifugation at 10,000 rpm for 15 min. Then, the pH of the sample was adjusted back to neutral (~pH 6.7) with 6 M NaOH prior to dilution with DI water. This method is modified from Jen et al. (14). Method 4, the urine sample was treated with acid, centrifugation and base as in Method 3, and also followed by filtration through a 0.22 µm syringe filter prior to dilution with DI water. Four urine samples were prepared in four different ways as described. Samples of all four methods were run with TLC. To avoid column clogging, only samples of Method 2–4 were also run with HPLC. According to the linear working ranges, the optimum dilutions of samples at 50 and five times were selected for HPLC and TLC, respectively. It was found that Method 1 (untreated urine) gave RGB readings that are significantly different from Method 2 (filtered urine), when comparing by the paired t-test at 95% confidence. Methods 2, 3 and 4 are not significantly different in both TLC and HPLC analyses. This indicates that untreated urine should not be used and that it is important to remove particulates and proteins from samples. Precipitation of protein with acid, followed by centrifugation and pH adjustment, was comparable to filtration with an MCE membrane, but involved more steps. To reduce the sample preparation time, one-step filtration of urine (Method 2) was selected for further experiments.
Another step that can cause error is the reading of RGB values. The image of the stained TLC plate was opened in the Microsoft Paint program. The color in the middle of the creatinine spot was picked using the “color picker” function, and then the RGB values of the selected color were revealed with the “edit colors” function. Although it is impossible to pick exactly the same pixel every time by clicking at the middle of the spot by aiming with bare eyes, the precision of the creatinine concentration resolved from the image analysis based on 10 replicated readings of green values of the same creatinine spot was found to be very satisfactory (2.03% RSD). This is because the middle areas of all the spots have an even intensity as proved by making the background color of the TLC plate transparent. This can be done in Microsoft Power Point using the “color” function located under the Format tab in Picture Tools. By placing the cursor on any part of the background on the TLC plate picture, and selecting the “Set the transparent color” function, all the pixels of that selected color become transparent. As can be seen in Figure Figure4,4, the background is made evenly transparent and the TLC spots and solvent front are shown clearly. Each spot shows an even color, which indicates that the color shade and intensity of any pixel in the middle of the spot can be used for the analysis and will result in a high-precision color reading. In addition, since the program cannot make more than one color become transparent, Figure Figure44 also proves that the background color is uniform because all the background color could be completely removed in one step. Therefore, it is unnecessary to subtract background color when analyzing the image of each spot. This is an improvement over the previous reports (10, 11) in which background color correction steps were needed due to uneven staining. Elimination of background correction helps to simplify the image analysis process as well as to improve precision of the analysis.
Accuracy of the method was also evaluated by studying percent recovery. Four urine samples were analyzed with the TLC as individual samples and as mixtures of two different samples at the ratio of 1 : 1 (0.5 mL each) or 1 : 2 (0.5 and 1 mL). Based on the concentration of each of the four individual samples (samples A, B, C and D in Table TableII)II) determined from the calibration graph of green values obtained from the series of creatinine standards analyzed on this same TLC plate, the expected concentrations of creatinine in the final volume of mixtures (M1–M8 in Table TableII)II) were calculated using Equation (1). These calculated concentrations of the mixtures were then compared with the concentrations of the mixtures obtained from the calibration graph. Excellent recovery of 94–102% was achieved, as shown in Table TableIIII.
where Cm is concentration of the mixture in mg/mL, C1 and C2 are the concentrations of creatinine found in each individual sample (mg/mL), and V1 and V2 are volumes in mL of the two samples that are used to form the mixture.
From the series of standard creatinine solutions (0.0016–2.3150 mg/mL) spotted on the TLC plate, the lowest concentration spot that can be observed by the eyes to have distinguishable color when compared with the background was 0.0810 mg/mL. Therefore, the limit of detection (LOD) was estimated to be equivalent to 0.24 µg of creatinine per 3 µL spot. Given that the limit of quantification (LOQ) is the lowest point of the linear working range, 0.0810 mg/mL is also the LOQ for green and blue values, while the LOQ value for red is 0.2315 mg/mL; see Figure Figure3D.3D. For a 20 × 10 cm TLC plate, a total of 19 spots (3 µL) can fit comfortably at a spacing of 1 cm. The number of samples that one TLC plate can accommodate per run depends on the number of standard solutions used. Here, seven standard concentrations were used. Therefore, a maximum of 12 urine samples can be run on the same TLC plate. To ensure that the calibration graphs can cover the concentration range in the sample solutions, at least five standard concentrations should be used, and urine samples should be diluted to the appropriate concentration. From several trials, it was found that two to five times dilution is the most appropriate for these urine samples.
Since the objective of this study is to compare the TLC-image analysis method with the HPLC method, it is important to have a sufficient number of samples to demonstrate the effectiveness of the method. Therefore, some volunteers were asked to collect urine samples twice, both the night before and the morning of the experiment. In addition, urine samples were randomly picked and spiked with either 10 or 50 µL of standard creatinine solution (1.1575 mg/mL) into a portion of 500 µL of urine. This was to create a total of 15 urine samples (some were nonspiked and some were spiked urine samples). Then, they were randomly labeled as S1–S15 and filtered through the MCE membrane. The filtered urine samples were diluted 50 times for HPLC analysis and five times for TLC analysis to obtain signals in the working ranges of those techniques. For HPLC isocratic elution of a 20 µL sample from the 10 cm long C18 column with 0.05 M ammonium phosphate buffer (pH 7.4) at the flow rate of 1 mL/min, the creatinine peak was eluted at tR ~2.40 min. Standard creatinine solutions used for construction of HPLC linear calibration graphs, based on peak area, were in the range of 0.0016–0.1620 mg/mL. The TLC linear calibration graphs obtained from green values of the standard creatinine were in the range of 0.0810–0.9260 mg/mL. After multiplying with the dilution factors, the concentrations of urinary creatinine were determined and compared, as shown in Table III. Concentrations determined from various staining times from 30 to 60 min were found to be comparable (using analysis of variance, ANOVA), which means that a staining time as short as 30 min can be used. Comparing results using a paired t-test at 95% confidence, TLC results obtained from each staining time from 30 up to 60 min are not significantly different from those obtained from HPLC.
Although TLC has lower separation efficiency than other instrument-based chromatography such as GC and HPLC, it is adequate for separation of samples containing a few key analytes. For example, TLC is cost-effective for analysis of pharmaceutical samples, especially in resource-limited countries (1). With the advancement of computer software and devices such as scanners and digital cameras, image analysis of the TLC plate satisfactorily enables quantitative determination of the analyte of interest with minimum training requirements. For analysis of organic molecules without chromophores, using staining reagents is a preferred method of choice (10). Most studies have used special TLC software for image analysis, which would contribute additional cost to the TLC process. This work aims to explore a lower-cost alternative by using commonly available software to quantify analyte from the image of the stained TLC plate through the RGB values.
Urinary creatinine was selected as a model analyte to demonstrate the efficiency of the method in analysis of a complicated body fluid sample. With a normal sample cleanup step, creatinine can be separated from other substances using a suitable mobile phase. Due to the nature of the TLC process in which the operator has more chances of exposure to volatile organic solvent mobile phases when compared with other instrument-based chromatography such as HPLC or GC, the authors emphasize the use of lower toxicity solvents as mobile phases. Therefore, even though shorter separation times have been reported using powerful mobile phases of high toxicity such as chloroform, acetonitrile, pyridine and formic acid (17, 18), the selection of mobile phases in this study only includes lower toxicity solvents such as propanol and butanol. Even though the separation time may be compromised, the main objective of this study in demonstrating and evaluating the proposed methodology is achieved with these less toxic mobile phases.
The staining process is critical and it is important to set up the staining chamber that provides an even stain for the whole TLC plate. Placing the TLC plate face down in a shallow staining chamber to expose it to I2 vapor from I2 crystals, which were spread to cover the bottom of the chamber, was found to be the most effective arrangement. Since the staining is completely uniform, the background removal step is not necessary because the background color affects the whole TLC plate equally. The RGB values can be obtained from any pixel in the middle of the separated spot with highly satisfactory precision. To ensure inter-day precision, the amount of I2 crystals is kept constant daily. The staining time should also be long enough to completely stain all the spots. From this study, the separation of urinary creatinine by butanol–NH4OH–H2O (40 : 10 : 50 v/v) using a 3 µL spot in the range of 0.0810–1.1575 mg/mL requires a staining time of at least 30 min. This should depend on the nature of the analyte, its concentration range, amount of I2 crystals, as well as the dimension of the staining chamber used.
The RGB color model is the expression of color based on the additive primary colors of light. Typically, the RGB color selector in the graphic software presents the RGB values in the range of 0–255. The zero value corresponds to the strongest intensity of that color (i.e., no light, or darkest color), while the highest value of 255 corresponds to the lowest intensity or the lightest shade of that color. Therefore, the spot with intense color (high analyte concentration) has low RGB values and vice versa. This is why the slopes of the graphs in Figure Figure33 are negative. It was found that the green value yields the most useful accurate analysis results from the brownish I2 staining color when compared with HPLC. If the staining reagent is in another color, it is important to re-evaluate whether the red, green or blue value is the most effective. It may also be possible that different analytes may have different chemistry with I2 vapor and give different color tone when compared with creatinine. Therefore, the use of red, green or blue should also be re-evaluated when studying different analyte species.
The proposed simple image analysis of a TLC plate based on staining with universal reagent (I2 vapor) and using RGB values for quantitative determination of urinary creatinine was successfully demonstrated. The RGB values from the commonly available software such as Microsoft Paint can be used as an alternative to more expensive tailor-made software. Meaningful results can be obtained with this low-cost TLC-digital image analysis setup. Although different staining times may be needed, similar procedures should be applicable for analysis of other substances.
This work is supported by the 2014 Summer Fellowship from the Faculty Development Program, and by the 2015 summer research fund through the College of Arts and Sciences, Xavier University.