INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

THE MICROSPHERE TECHNIQUE is still considered the "gold standard" for measuring regional blood flow, either as an absolute value or relative to other organ flow (4, 13). Until recently, only radioactive spheres were used, but the emergence of colored microspheres provides the opportunity for a simple and safe technique, particularly by avoiding the use of radioactivity (9, 12). Briefly, the technique is as follows. A known number of microspheres is injected in a mixing chamber such as the left atria or the left ventricle, and a reference arterial sample is drawn at a known withdrawal rate. After animals are euthanized, the number of spheres is measured both in the reference sample and in the sample organ of interest. Regional blood flow is then calculated relative to the reference withdrawal rate. Thus errors in both measurements will participate in the error in regional blood flow calculation. In fact, measurements of regional blood flow by microspheres suffer from lack of precision due to distributional variation governed by the Poisson distribution and due to assay imprecision (2, 5, 6). The assay error is most often neglected in case of radioactive microspheres because increasing counting time meets the required precision (2). However, in the case of colored microspheres, the assay error may be the primary cause of imprecision (see RESULTS). One may thus be tempted to increase the number of microspheres injected. However, increasing the number of spheres in both tissue and reference samples will not decrease the error because the assay generally has a constant coefficient of variation (CV). Moreover, if one increases the number of spheres injected per measurement in the rat, serious hemodynamic disturbances are expected to occur (11). It is, then, of utmost importance to reduce the assay error when one uses colored microspheres. In that respect, the usual direct spectrophotometric method derived from radioactivity measurement lacks precision (see RESULTS). If some studies have addressed the question of recovery of dye (8), none (in our knowledge) has studied the problem of reproducibility of the assay.

We now propose 1) the use of an internal standard technique and 2) a new assay using high-performance liquid chromatography (HPLC), which we compare with the direct spectrophotometric assay in terms of reproducibility, i.e., relative error.

	METHODS

Top Abstract Introduction Methods Results Discussion References

Materials and reagents. Colored (white, yellow, violet, red, and blue) microspheres (DyeTrak Triton 15.5 µm) were purchased from Bioseb (Paris, France). Inasmuch as the whole study is based on the principle of the reference sample, we used the particle concentration recorded on the manufacturer's chart (3,000,000 spheres/ml except for the violet spheres, whose concentration was recorded to be 3,300,000 spheres/ml) with no further verification. N,N,-dimethylformamide (DMF) (HPLC grade) and polyoxyethylenesorbitan monooleate (Tween 80) were purchased from Sigma (St. Quentin Fallavier, France). Acetonitrile (HPLC grade), ethanol (low grade with 4% methanol; a preliminary experiment has shown that the quality of extraction and background noise were not modified whether HPLC-grade or low-grade ethanol was used), NaCl, NaH₂PO₄, and H₃PO₄ were purchased from Prolabo (Paris, France).

Chromatography. Chromatographic analysis was performed under isocratic conditions at room temperature. The HPLC system consisted of a Shimadzu LC9A pump (Touzart et Matignon, Les Ulis, France), a variable-wavelength ultraviolet (UV)-visible detector (model 1050, Hewlett-Packard, Les Ulis, France), and a Linseis L 6512B pen recorder (Bioblock, Illkirch, France). The separation was performed on a Waters Resolve 5-µm spherical C₁₈, 3.9 × 300 mm column (Waters, St. Quentin en Yvelines, France). The mobile phase consisted of 35% 0.01 M NaH₂PO₄ buffer containing 0.001 M sodium dodecyl sulfate and 65% acetonitrile (vol/vol) and was set at a flow rate of 1.2 ml/min. Twenty-five microliters of the final extract were injected in the system. Detection was made at wavelengths of 370, 448, 268, and 530 nm, respectively, for the white, yellow, violet, and red dye. The blue spheres were not used because a preliminary experiment showed that a marked degradation of the dye occurred during extraction with the appearance of multiple peaks after chromatographic separation.

View larger version (20K): [in this window] [in a new window]   Fig. 1.   Top: chromatograms obtained after extraction of 0.5 ml blood containing 25 yellow (2), 50 white (1), and 50 red spheres (4) with 1,500 violet spheres (3) as internal standard (IS) (A); 125 yellow (2), 250 white (1), and 250 red spheres (4) with 1,500 violet spheres (3) as IS (B); or 1,000 yellow (2), 2,000 white (1), and 2,000 red spheres (4) with 1,500 violet spheres (3) as IS (C). Bottom: chromatograms obtained after extraction of rat brain samples with the following measured content: 27 white (1), 11 yellow (2), and 20 red spheres (4) with 1,500 violet spheres (3) as IS (D) and 451 white (1), 445 yellow (2), and 335 red (4) spheres with 1,500 violet spheres (3) as IS (E). In D, U is unknown peak.

Spectrophotometric procedure. Sample spectra were recorded using the same variable-wavelength UV-visible detector as used for HPLC, set in the scanning mode. One hundred to one hundred twenty microliters of the extract were injected using a direct injection port with minimal dead volume (Rheodyne, Touzart et Matignon, Les Ulis, France). Absorption was measured at 335, 375, 405, 435, 475, 505, 535, 565, and 595 nm. Dye concentration was determined using a modification of the procedure described by Schosser et al. (15). This modification allows the use of a number of wavelength measurements greater than the number of unknown dye concentrations by using linear least-squares regression using the Faddeev-Leverrier algorithm (10) rather than a simple matrix inversion procedure. Because background noise in biological samples is greater in the lower wavelength spectrum (typically at wavelengths between 190 and 400 nm), we compared the results obtained using the set of nine wavelengths to the results obtained using a reduced set of seven wavelengths (from 405 to 595 nm) and to the results obtained with the classical four-point technique [number of measurements equal to number of colors (405, 475, 535, and 595 nm)].

Extraction. The extraction procedure was always performed using 1% Tween 80. The spheres were gently vortex-mixed for 30-60 s (at low speed to be sure that no foam was formed), then put in an ultrasonic bath and hand-mixed just before sampling with a micropipette. Direct extraction of dye from the spheres was performed using DMF and was considered complete. Extraction from biological material was performed as follows. Heparinized blood (500 µl) or 200- to 300-mg tissue samples (heart, kidney, liver, or brain) obtained from Sprague-Dawley rats were spiked with known amounts of spheres in 5-ml polypropylene tubes. The spheres were introduced into the tubes by pipetting known numbers of spheres (in 50 µl saline with 1% Tween 80). In spiked samples to which a known number of violet spheres were added, this number served as an internal standard reference. For the lower concentrations tested (50 yellow, 100 white, and 100 red spheres) and for extraction after in vivo injection, 1,500 violet spheres were added to the sample as internal standard. Two milliliters of distilled water with 1% Tween 80 were added, and the tubes were hand-shaken. One milliliter of KOH (12 M) was added, and the tubes were put in a water bath shaker at 70°C for 2 h 30 min. After being cooled, the tubes were centrifuged at 2,500-3,000 revolutions/min for 20 min. The lower part was carefully aspired with a thin needle and discarded, leaving the spheres in the upper layer (0.5-0.75 ml). Four milliliters of ethanol containing 1% Tween 80 and 100 µl H₃PO₄ (6 M) were added, and the tubes were vortex-mixed for 30-40 s and ultrasonicated for 4-5 min. After centrifugation at 2,500-3,000 revolutions/min for 10 min, the supernatant was discarded and the procedure was repeated. The residue was then evaporated to dryness under vacuum at 35°C for 40 min. The dry residue was extracted with 200 µl DMF, vortex-mixed for 30-40 s, and ultrasonicated for 4-5 min. After centrifugation for 10 min at 4,000 revolutions/min, 25 µl of the supernatant were injected into the HPLC system or 100-120 µl were directly injected in the spectroscopic cell for measurement by the traditional technique.

Protocol design. Experiments were conducted using five replicates. First, we assessed the pipetting variability 1) by pipetting 50 µl of solutions containing 4,000, 400, and 40 white microspheres (these concentrations were obtained by serial dilutions) into test tubes and 2) by pipetting 150,000 white microspheres (50 µl) in a polyethylene tube (PE-50, Guerbet Louvres, France) using a model 44 Harvard pump (Harvard, Les Ulis, France) and rinsing the tube with 2 ml water containing 1% Tween 80. Because each dye did not display the same absorbance intensity, we normalized the mixtures of spheres according to their respective absorbance, i.e., to the intensity of their respective signal. Recovery and reproducibility of extraction and detection in biological samples were tested using three different mixtures: 1) 50 yellow, 100 white, 100 red, and 200 violet spheres (violet spheres were not added at this time when an internal standard procedure was used); 2) 400 yellow, 800 white, 800 red, and 1,600 violet spheres; and 3) 1,500 yellow, 3,000 white, 3,000 red, and 6,000 violet spheres.

In vivo experiments. Two Sprague-Dawley rats weighing 380-400 g were anesthetized with pentobarbital (50 mg/kg ip). The left femoral artery was cannulated using a 22-gauge Teflon short catheter (Insyte, Beckon-Dickinson, Dun Laoghaire, UK), and a polyethylene catheter (PE-50, Guerbet Louvres, France) was advanced in the left ventricle via the right carotid artery for dye injection (8). Rat 1 received a mixture of 300,000 white, 300,000 yellow, and 300,000 red spheres over a 40-s period using a model 22 Harvard pump (Harvard, Les Ulis, France), and rat 2 received a mixture of 100,000 white, 100,000 yellow, and 100,000 red spheres. The animals were euthanized with an overdose of pentobarbital, and the organs (heart, brain, and right and left kidneys) were dissected and cut into five samples weighing 150-300 mg each (200-400 mg for brain tissue). After extraction as described above and elution with 200 µl DMF, 25 µl were injected into the HPLC system and then 100-120 µl were injected directly into the spectrophotometer.

View larger version (16K): [in this window] [in a new window]   Fig. 2.   Measured vs. spiked dye concentration obtained after fitting an 8-point standard curve (7 points for violet) (A), and the corresponding weighted residuals (B). Weighting as the inverse of concentration squared led to small bias.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

HPLC. The separation was optimal with mobile phase containing 65% acetonitrile (vol/vol) (Fig. 1). Under the above-described conditions, retention times were as follows: white spheres, 5.1 min (CV 1.3%); yellow spheres, 10.6 min (CV 1.0%); violet spheres, 12.6 min (CV 0.7%); and red spheres, 13.2 min (CV 0.7%). The lower limits of detection of the dye (directly extracted from the spheres) at four times the size of the basal noise were 2.5 spheres (white), <1 sphere (yellow), 12.5 spheres (violet), and 2.5 spheres (red). The CV within the same extracted sample at four times the lower limit of detection were 0.6% (white spheres), 1.0% (yellow spheres), 1.7% (violet spheres), and 3% (red spheres). The errors in pipetting were, respectively, 4, 7, and 15% for the solutions containing 4,000, 400, and 40 spheres. The error in pipetting 50 µl of spheres in PE tubing was 4%. An eight-point calibration curve was constructed in blood (7 points for violet spheres). The curves were found to be linear from 20 to 5,000 spheres (yellow, r = 0.998), from 40 to 10,000 spheres (white, r = 0.996; red, r = 0.995), and from 80 to 10,000 spheres (violet, r = 0.993) (Fig. 2). However, recovery of dye was incomplete even at sphere concentrations 400 per sample (Table 1). Reproducibility was poor, but when violet spheres were used as internal standard, reproducibility markedly increased with a CV between 1.8 and 5.1% at the higher concentration and between 4.1 and 18% at the lower concentration, depending on the type of spheres and type of tissue (Table 1).

Direct spectrophotometry. In contrast to the above results concerning HPLC, the classical direct spectrophotometric method showed very poor reproducibility, and the difference between classical spectrophotometric and HPLC methods using internal standards was highly significant (Table 2). After scanning the samples, the calculations used linear least squares with a nine-point method (from 335 to 595 nm), a seven-point method (obtained by deleting the first 2 points, at 335 and 375 nm, because the high background noise observed at these wavelengths was supposed to decrease reproducibility), or the classical four-point method (number of measurements equal to number of unknowns). We observed an improvement in the quality of results when the reduced seven-point method was used compared with the nine-point technique. In contrast, the four-point technique always showed an impairment in the quality of results (except in the case of white spheres, for which the 3 techniques gave similar results). It was not possible to measure dye content at sphere concentrations less than 200 (yellow), 400 (white and red), or 800 spheres per milliliter (violet): CVs and bias both always exceeded 60% at these concentrations.

In vivo comparison of the two techniques. The results of the in vivo experiment are summarized in Table 3. With the exception of brain measurements in rat 2 (which received only 100,000 spheres), the two methods gave similar mean results, but HPLC always showed lower CVs than the traditional spectrophotometric technique using internal standard. However, the variability within an organ was still greater than the in vitro variability.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

The present study shows that assaying dye content of colored microspheres by HPLC with internal standard is a major improvement compared with the direct spectrophotometric assay. The extraction step needs to be carefully performed. In fact, use of surfactant is critical. Some authors use filtration in the washing-extraction process (12). This may improve the sample washing procedure but with the risk of losing spheres. We made the choice of working in a unique tube. This procedure is assumed to lose fewer spheres but may result in incomplete clean-up procedure. In fact, recovery was incomplete at all particle concentrations, especially in blood and in the liver, as already described by Hakkinen et al. (8). However, these authors used a much larger number of spheres in their samples (5,000 spheres) compared with the range of spheres used in our experiment.

The lower limit of detection of the HPLC technique was far below the number of spheres usually encountered in tissue or blood samples. When there are 400 microspheres in the sample, the relative error of the assay is then between 2 and 10% with the yellow, white, and red microspheres. To date, only three different measurements are made possible because, of the five available colors, one served as internal standard (violet) and another (blue) exhibited marked degradation during the extraction or separation procedure, leading to the unexpected observation of three different peaks. We chose violet as standard because its absorbance is much lower than the absorbance observed with the other colors, thus leading to a greater CV at the critical value of 400 spheres/sample: ~1,500 spheres appears then to be the adequate number to be added as internal standard (see Fig. 1). The use of the internal standard markedly improved the precision of the technique (Table 1). This is not surprising because recovery appears incomplete even at high particle concentrations. It should also be noted that we only measured recovery relative to dye directly extracted from spheres in DMF and not relative to the pure dye molecule. The technique of isocratic HPLC with UV-visible spectrophotometric detection is simple and easy to implement. However, it is certainly more time consuming than the direct spectrophotometric method; the use of an autosampler is highly recommended.

At the critical value of 400 spheres/sample, the direct spectrophotometric method cannot be recommended because of unacceptable error in flow measurement due to lack of precision of the assay. The high background noise associated with the direct spectroscopic measurement method is responsible for this imprecision. In contrast, chromatography adds a separation step that markedly increases the signal-to-noise ratio and permits a single determination of the compound of interest.

In vivo injection of spheres in rats confirms these findings. The mean number of spheres measured in samples by the two methods is not markedly different. This confirms the results of the in vitro experiments showing that bias was (on average) not very important. Nevertheless, when the number of spheres in the samples was low (<300 in brain of rat 2), a systematic difference in the number of spheres measured by the two techniques became visible. The average intraorgan CV was three times greater with traditional spectrophotometry than with HPLC (Table 3). In fact, physiological variability in sphere deposition in organs is well documented. For example, Bassingthwaighte et al. (3) showed that relative dispersion of flows measured in baboon, sheep, and rabbits using spheres or soluble markers was 23-30% (mean of 6-11 animals). Our results of single-organ variability in sphere content (measured with HPLC) is consistent with these findings. Nevertheless, the combined error made in measuring reference blood and tissue contents is expected to increase the resulting CV of regional blood flow.

Cost and time requirements. The extraction procedure is the same for the two techniques. However, these two techniques differ in equipment, in time needed for spectrophotometry (traditional or after chromatographic separation), and in data processing requirements. The classical spectrophotometric technique requires the use of a spectrophotometer with an automatic scanning mode, leading either to a total spectrum or to the measurement of absorbance at predetermined wavelengths, usually coupled with a computerized recording system, whereas HPLC needs a simpler spectrophotometer (with the same precision and sensitivity). After the spectroscopic procedure, a computer program is needed for inverting the matrix of absorbance. On the other hand, HPLC requires the use of other equipments (pump, injection port, and column) and also the use of solvents. Moreover, HPLC is time consuming, and the use of an automatic sampler is highly recommended. The counting unit and computer software used for traditional spectrophotometry cost approximately $18,000, whereas an HPLC system costs approximately $17,000-$20,500 (isocratic pump $6,000, injection port and tubing $1,500, UV-visible detector $8,000, and pen recorder $1,500 or integrator $5,000). The replacement of the column ($500 for ~1,000 samples, i.e., $0.50 per sample assayed) and the extra cost of solvents ($30 for ~100 samples, considering mobile-phase recycling) have also to be taken into account. Also, there is an extra cost due to the time required for chromatographic separation. This cost is about $6-$8 per sample, considering $25 per hour for personnel costs. The addition of an automatic sampler to the HPLC system is highly recommended. Its cost ($6,000-$8,000) is then paid off after ~1,000 samples are assayed. Finally, the use of HPLC adds an extra cost of about $5,000-$8,000 for equipment and $0.80 per sample assayed compared with the traditional spectrophotometric technique. However, many laboratories already possess an HPLC system that can be used for different purposes.