Tumor hypoxia is a major barrier to tumor radiation therapy. Typically tumor hypoxia occurs in two forms: chronic and acute. Although the existence of acute hypoxia has long been acknowledged, its temporal characteristics have never been directly measured and documented. In this study tumor , blood flow (BF), and arterial blood pressure (BP) were measured simultaneously in nine Fischer 344 rats bearing R3230Ac rat mammary adenocarcinomas in the subcutis of the left hindleg. We measured at a single location for 36–125 min using recessed-tip oxygen microelectrodes. Simultaneously, we measured tumor BF at two sites within the tumor using laser-Doppler flowmetry (LDF). Similar recordings were made in the quadriceps muscle of seven non-tumor-bearing rats. The , tumor BF, and BP records were subjected to Fourier analysis. and BF showed low-frequency fluctuations (<2 cycles/min) in both tumor and muscle, but the magnitude of the changes in tumor was greater. Tumor BF showed more activity at low frequencies than muscle BF, and the magnitude tended to be greater. No strong correlations were found between and BF power spectra for either tumor or muscle or between the frequency patterns of BP and tumor spectra. These results quantitatively demonstrate, for the first time, that BF and fluctuate at very low frequencies in tumors. In addition to having biological significance for tumor therapy, these fluctuations may have the potential to alter tumor cell behavior via induction of hypoxia reoxygenation injury and/or altered gene expression.
- oxygen tension
- spectral analysis
the presence of hypoxia in tumors has long been known to adversely affect the sensitivity of tumors to radiation therapy (44). More recently hypoxia has been shown to exert a selective pressure for the survival of those cells with lower apoptotic potential (15), to alter gene expression such as those involved in cell cycle regulation and cytokine production (10), to increase mutation frequency (35), and to impact treatment outcome and patient survival (3, 19, 20,32, 33). Significant portions of both human and animal tumors have been shown to be hypoxic by direct measurement with oxygen electrodes (2,46), optical oxygen sensors (9), and optical fluorescent techniques (17) or by use of hypoxia markers (24).
Despite the importance and prevalence of tumor hypoxia, its exact origins have still not been totally elucidated. Typically hypoxia is divided into two categories: “chronic” and “perfusion limited” (or acute). Chronically hypoxic cells are cells in a low-oxygen environment created largely by their distance from the blood vessels and the metabolic activity of the intervening cells (44). The second category of hypoxic cells, the acutely or transiently hypoxic cells, experience low as a result of transient changes in the oxygen supply (4). Classically this “perfusion-limited” hypoxia has been attributed to transient blockages in the vessels or collapse of tumor vessels, leading to temporary cessation of blood flow (BF) or intermittent flow (33).
The term intermittent flow typically implies that BF in the vessel temporarily shuts down completely, causing transient tissue hypoxia. When the vessel reopens, cells dependent on that vessel for oxygen are reoxygenated. Whereas intermittent flow definitely occurs in tumors (31) and probably results in intermittent tumor hypoxia, another subtler possibility for intermittent hypoxia may be fluctuations in red blood cell (RBC) flux through a vessel. Large fluctuations in RBC flux and perivascular occur in R3230Ac mammary carcinoma implanted in the dorsal window chamber of rats (12,26). Similar heterogeneity in RBC flux has also been demonstrated in other solid tumor models (6, 18) and in human tumors in patients (18) as measured by laser-Doppler flowmetry (LDF). It has been theoretically demonstrated that such fluctuations in RBC flux can have significant effects on tumor interstitial and the hypoxic fraction of tumors (26).
Although changes in RBC flux and perivascular have been demonstrated in vivo, the time course of changes in parenchymal in a solid tumor has never been measured. The hypothesis of the current study was that interstitial would fluctuate and that this fluctuation would differ from that seen in a normal tissue such as muscle. To address this hypothesis, R3230Ac mammary adenocarcinomas were implanted into the hindlimbs of Fischer 344 rats. We measured at a single site in 1-cm-diameter tumors for 36–125 min using recessed-tip oxygen microelectrodes. Tumor BF was measured simultaneously using LDF, and arterial blood pressure (BP) was monitored as well. Similar experiments were performed in the hindlimb muscle of rats. To quantitatively characterize the fluctuations, the Fourier power spectra of the and LDF recordings in tumor and muscle were determined. A portion of these data was qualitatively presented in an earlier study (11), but Fourier analysis, which is the emphasis of this study, was not included.
Sixteen female Fischer 344 rats (Charles River Laboratories, Raleigh, NC) were used in this study. Nine rats received subcutaneous implants of 1- to 2-mm3 pieces of R3230Ac rat mammary adenocarcinoma in the left hindlimb. After the tumor had reached 1 cm in diameter, the rat was anesthetized with an intraperitoneal injection of 50 mg/kg pentobarbital sodium. The femoral artery and vein were cannulated for later recording of BP and for venous access, respectively. A small portion (∼4–10 mm2) of the skin and tumor capsule was then removed to expose the surface of the tumor. This was kept moist by topical application of saline. We made a small incision in the left forelimb, and we sutured an Ag/AgCl reference electrode into the subcutis. Rat body temperature was maintained by placing the rat on a regulated water-heated blanket (K-Module, Baxter Healthcare, Valencia, CA).
Muscle and BF were measured in seven non-tumor-bearing Fischer 344 rats. We exposed the left quadriceps muscle by removing a patch of skin on the left thigh to permit insertion of the oxygen microelectrode later.
Recessed-tip oxygen microelectrodes were produced using a previously published technique (28). The resultant electrodes had tip diameters of 9.4 ± 3.1 μm (means ± SD; 16 electrodes). Recess lengths were relatively short, on the order of 30 μm. Several electrodes were used for more than one experiment.
The microelectrodes were polarized at −0.7 V using a commercial polarizing box and picoammeter unit (Chemical Microsensor no. 1201, Diamond General, Ann Arbor, MI). Electrodes were calibrated before and after experiments in a saline-filled tonometer alternately bubbled with 0%, 5%, 15%, or 21% oxygen (balance nitrogen). The saline was warmed to 37°C. An in vivo dead value was also obtained by recording the microelectrode current in the tissue after euthanasia of the rat with an overdose of pentobarbital sodium. The average sensitivity of the microelectrodes used in this study was 1.27 ± 0.72 mmHg/pA (means ± SD; 16 electrodes).
After preparation of the animal and calibration of the electrode, the rat was placed on a water-heated blanket, and the left leg was stabilized on a rubber pedestal with tape. Care was taken not to elevate the leg. The arterial cannula was connected to a blood pressure transducer and amplifier (model 11-G4143–01; Gould Instruments, Valley View, OH), and the amplifier signal was digitized at 25 Hz and recorded using data acquisition software (AT-CODAS; Windaq, DATAQ Instruments, Akron, OH). Two laser-Doppler flow probes (outer diameter, 480 μm) were inserted into the tumor on the side opposite from the exposed surface. The probes were connected to the flowmeter and data acquisition system (Oxford Array, Oxford Optronix, Oxford, UK). This commercially available system acquires data at a fixed rate of 20 Hz.
A micromanipulator (model MO102E, Narishige, Narishige, Japan) was positioned so that a dummy electrode could reach the exposed tumor or muscle surface, which was covered by a drop of saline. The actual microelectrode was then placed in the micromanipulator and advanced into this saline droplet. We allowed the electrode to polarize for several minutes in the saline before we advanced it into the tumor or muscle. The microelectrode was moved several millimeters into the tissue until a clearly nonzero current was obtained, so that the had an opportunity to either increase or decrease. This was particularly necessary in the tumor, which had many very hypoxic (even anoxic) areas. Because the goal of this study was to look at fluctuations in and examine them as indicators of transient hypoxia, we did not want to leave an electrode in an area that may have been chronically hypoxic with readings near zero. The signal from the microelectrode was sent to the same data acquisition board and software as the BP (AT-CODAS) and digitized at 25 Hz.
BP, , and BF were recorded continuously for up to 125 min. If the mean arterial BP (MABP) dropped below 80 mmHg, the experiment was terminated early. At the end of the recording time, an overdose of pentobarbital sodium was injected intravenously. The recordings of all parameters continued for at least 5 min after death. The value after death was used in the final electrode calibration as a true in vivo zero as described previously (14). The BF values from the LDF signal recorded after death were used as the zero value, and the values were subtracted from the recorded values to produce a corrected BF. Any probes that did not show a decrease in BF after death were excluded from the analysis.
All Fourier analysis was performed using commercial software included in the data acquisition package (CODAS, DATAQ Instruments, Akron, OH). BP and recordings could be analyzed directly within the program, because that same software was used to acquire the data. The BF data had to be transferred from the Oxford array system to multiple ASCII text files, which were subsequently combined and converted to CODAS files.
The frequency characteristics of the instrumentation had no impact on the Fourier analysis. The oxygen microelectrodes respond within 40–100 ms (39), which is much faster than the fluctuations of interest in this study. Although the high-frequency response of the Chemical Microsensor no. 1201 is somewhat limited in the range (200 pA full scale) that was used in most cases, its cutoff (3 dB) frequency was measured to be 2.3 Hz (138 cycles/min), and the attenuation was 50% at 4.1 Hz (246 cycles/min). According to the manufacturer, the Oxford laser-Doppler array has a cutoff frequency of 5 Hz (300 cycles/min). Because these cutoff frequencies are relatively high, and the resulting attenuations had little or no effect on the apparent power within the frequency range of interest, the instrumentation should not have contributed to or biased the and LDF signals.
Two different record lengths were analyzed using the Fourier analysis: 5.5- to 6.8-min and 32.8- to 34.1-min records. Consecutive data points (8,192) recorded at either 25 or 20 Hz constituted the short-term records (5.5 min for and BP; 6.8 min for BF). Each record overlapped the neighboring records by 50%. For example, the second short-term Fourier analysis of a recording covered data from 2.73 to 8.19 min, whereas the third covered 5.46 to 10.92 min. The long-term (33 min for and BP; 34 min for BF) analysis was performed on records that contained 49,146 points (BP and ) or 40,960 points (BF). The records were averaged, over every 5 or 6 points by the software, to obtain files 8,192 points in length. Again each record overlapped the neighboring records by 50%. For example, the second long-term Fourier analysis of a recording covered data from 16.38 to 49.15 min, whereas the third analysis covered 32.76 to 65.53 min.
Each type of analysis yielded different information. Obviously many 5.5–6.8 min spectra could be generated for each recording, so the time dependence of the fluctuations could be easily shown using this type of analysis. Unfortunately, this information was gained at the cost of frequency resolution, because fewer points were included in the analysis. Resolution of the low-frequency range was obtained by analyzing 33- to 34-min segments of the recordings.
An informative way to view and compare frequency information from Fourier power spectra is to examine the frequencies at which the power peaks. This technique was recently used by Buerk and Riva (5) to compare power spectra of nitric oxide and BF in the optic nerve head. In this study, the analysis was used as follows: the frequencies at which the first 20 power peaks occurred were determined for each power spectrum. We then pooled peak frequencies for each tissue type, and we plotted the peak frequencies against each other. Differences in the dominant frequencies between tissues appeared as a deviation of the points from the identity line.
Power and magnitude of fluctuations.
The power of the Fourier power spectra (i.e., the magnitude of the fluctuations) was analyzed in two different ways. The simplest method of comparing power between spectra is to look at the total power over a given frequency range (1). Because most of the power in the present study was at very low frequencies, the range chosen for analysis was 0–10 cycles/min. The total power of a spectrum was obtained by summation of all powers between 0 cycles/min and a given frequency. The total cumulative power was plotted as a function of frequency, and comparisons were made. The total power between 0 and 10 cycles/min was also calculated, and this parameter was designated the total low-frequency power (TLFP).
Another way to examine the magnitude of the fluctuations is to determine the magnitudes of the peak powers in the spectra and compare these values among groups. This technique gives more detailed information about the power at each frequency rather than in a given range. In this type of analysis, the power at each peak of the power spectrum was determined and plotted as a function of the peak number. These peak powers were then pooled for each tissue type and plotted against each other. Differences in the peak powers between tissues appeared as a deviation of the points from the identity line.
All data were compared using nonparametric analysis. Differences among groups were tested using the Mann-WhitneyU-test. Differences among paired data were tested using the Wilcoxon’s rank sum test. Significance was achieved if P < 0.05.
Tumor was measured at 13 different sites in 9 tumors. The duration of the measurement period ranged from 36 to 125 min with means ± SD of 84 ± 22 min. The tumors were generally very hypoxic, and the electrode needed to be moved along the insertion track to find a location where a nonzero could be recorded. The average during the first 10 min of recording was 15.2 ± 10.5 mmHg (means ± SD;n = 13) with a range of 1.3 to 38.5 mmHg and a median of 12.9 mmHg. Overall there was a tendency for to decrease during the measurement period. At the end of the recording time, the average over 10 min was 6.9 ± 9.3 mmHg (means ± SD; n = 13) with a range of 0 to 33.7 mmHg and a median of 3.7 mmHg. The difference between the starting and ending was significant (P < 0.01; Wilcoxon test). This point will be addressed in thediscussion.
An example of one tumor recording is shown in Fig.1 A. In this particular example the fluctuated between 0 and 15 mmHg during a 77-min period. Examination of the trace reveals that the magnitude of the fluctuations changed over time. For example, the change in between 8 and 16 min was ∼9 mmHg, whereas the magnitude of the fluctuations between 24 and 30 min was only 2–3 mmHg. This is an experiment in which the fluctuations were large and of a somewhat regular pattern. A second example is shown in Fig.2 A. In this particular experiment, the fluctuations were not as regular. The fluctuated slowly from 12 to 8 to 15 mmHg during a 98-min recording period.
Fourier analysis of tumor fluctuations.
Fourier analysis of the tumor recordings showed that the fluctuations typically occurred at very low frequencies (<1 cycle/min) and were relatively large in magnitude. As noted in methods, two different analyses were carried out: a short-term (5.5 min) and a long-term (33 min) analysis. In Fig. 1, B andC, and Fig. 2,B andC, both types of analysis are shown. In Fig. 1 B and Fig.2 B the results of the short-term Fourier analysis are shown, whereas Fig.1 C and Fig.2 C show the long-term analysis.
The 5.5-min spectra reveal the changes in the magnitude and spectral characteristics of the fluctuations over time. This is best shown in the spectra for rat 1 (Fig.1 B). In the first 15 min of the recording period, the peak power was ∼15 mmHg2 and all of the activity was below 0.732 cycles/min. For the next 15 min, the power or magnitude of the fluctuations decreased considerably. The power then increased for the next 20 min, and some higher-frequency activity was evident at 1 cycle/min and beyond. This was followed by another short period of decreased activity, which was then followed by a return to larger fluctuations. In the second example, there were variations in short-term spectra over time as well (Fig.2 B). The overall magnitude of the fluctuations was much lower, and the spectra were typically dominated by the lowest frequency component (0.18 cycles/min). For ∼35 min, the fluctuations were small in amplitude but were spread out to frequencies beyond 1.5 cycles/min. For a period of 20 min, there was then little activity at all. This was followed by a period where the fluctuations were relatively large but were all at very low frequencies.
From each short-term spectrum, a total power could be calculated, which is an indicator of the overall magnitude of the fluctuations. Because the fluctuations were very slow, the pertinent information was in the low-frequency range (e.g., 0–10 cycles/min). Therefore, the TLFP (0–10 cycles/min) of each short-term spectrum was calculated. For example, the TLFP of the final spectrum (74 min) in Fig.1 B was 38.4 mmHg2. The average short-term TLFP value for the recording inrat 1 was 20.5 ± 8.9 mmHg2 (means ± SD,n = 27) with a range of 4.6–38.4 mmHg2 and a median of 20.3 mmHg2. The TLFP values for all of the short-term spectra analyzed in tumors are summarized in Table1.
The 33-min spectra show some of the low-frequency information that is not obtainable in the short-term analysis. Figure1 C shows the power spectra for three consecutive, overlapping 33-min recording periods from the tumor recording in Fig.1 A. From these spectra, it is apparent that not only the magnitude of the fluctuations can change, but the dominant frequency at which the fluctuations occur can change as well. In this example, the dominant frequency changed from 0.22 to 0.12 to 0.28 cycles/min during the three recording periods, although the power peak at those three frequencies was always ∼5.5 mmHg2. Figure2 C presents four consecutive long-term power spectra for the tumor recording of Fig. 2 A. This example is different from the other, in that all the spectra are dominated by the slowest frequency (0.03 cycles/min). The magnitude of the peak varies from 0.14 to 2.75 to 0.84 to 4.7 mmHg2. The second and fourth 33-min periods are dominated by the large, slow, and gradual changes in (Fig.2 A).
As was the case with the short-term analysis, the results of the long-term spectral analysis can be summarized by looking at the TLFP values. The TLFP values from the three spectra from the tumor inrat 1 were 19.3, 22.4, and 27.4 mmHg2. In rat 5b, the four TLFP values were 1.0, 3.7, 1.3, and 8.2 mmHg2. The long-term TLFP values for all 13 of the measured tumor sites are summarized in Table 1.
Tumor BF Fluctuations
Tumor BF recordings.
In addition to , tumor BF was simultaneously measured in the same nine rats at 21 different sites. In general, all of the probes showed some fluctuation in tumor BF. Figure3 shows one of the two traces of tumor BF recorded in rat 1 over the same 77 min as the in Fig.1 A. This particular LDF probe showed many large, slow changes in tumor BF of ∼10–30% (Fig.3 A). At some time points, tumor BF dropped ∼40% compared with the average baseline and then rebounded. The other probe in this tumor also showed large fluctuations in tumor BF (not shown). Whereas some of the characteristics were similar, the temporal characteristics of the two recordings made at different sites in the same tumor were very different. Tumor BF also fluctuated inrat 5b, but at a smaller amplitude (not shown). Both recordings in that particular tumor were dominated by a slow 20–40% increase at 55 min.
Fourier Analysis of Tumor Blood Flow Fluctuations
As was the case for the recordings, two Fourier transform analyses were performed on the tumor BF recordings: short term and long term. The results of the short-term (6.8 min) Fourier analyses of one of the tumor BF recordings inrat 1 are summarized in Fig.3 B. Most of the significant power was below 2 cycles/min. The LDF probe recorded tumor BF fluctuations with peak power values under 100%2 for the first 55 min of the recording. This would translate into amplitude changes of ∼10% or minimum-maximum changes of ∼20%, which is similar to what is seen in the original tumor BF trace (Fig.3 A). For the last 20 min of the recording period, the probe showed an increase in the peak power values of 100–400%2, or amplitude changes of ∼10–20%. Qualitatively, the first LDF probe (Fig.3 B) recorded a gradual increase in the magnitude of tumor BF fluctuations, whereas the second probe (not shown) detected fluctuations of relatively similar magnitude throughout most of the recording time.
The characteristics of the short-term power spectra are best summarized by examining the average TLFP values. The average short-term TLFP value for the first tumor BF recording in rat 1 (Fig. 3) was 253.7 ± 181.0%2 (means ± SD,n = 21) with a range of 23.3–654.8%2 and a median of 189.9%2. The average short-term TLFP value for the second tumor BF recording in rat 1 (not shown) was 248.9 ± 88.3%2 (means ± SD,n = 21) with a range of 162.8–474.5%2 and a median of 219.0%2. These values, particularly the standard deviation, show how different these traces were. The TLFP values for all of the short-term tumor BF spectra analyzed in tumors are summarized in Table2.
As was the case for the recordings, more detailed information on the nature of the tumor BF fluctuations can be gained by looking at the long-term power spectra. Figure 3 C shows the power spectra for three consecutive, overlapping 34-min recording periods from the tumor BF recording in Fig. 3 A. In this example, there was significant activity at frequencies out to and beyond 2 cycles/min. The dominant frequency remained relatively stable at 0.23 or 0.17 cycles/min, although the maximum power at those frequencies increased in the third 34-min period (Fig.3 C). In contrast, in the tumor BF recording of the second LDF probe, the dominant frequency changed from 0.56 to 0.03 to 0.64 cycles/min during the three recording periods, and the maximum power at those three frequencies was 44, 43, and 21%2, respectively (not shown). Thus the nature of the fluctuations measured by these two probes was very different.
Again, the results of the long-term spectral analysis can be summarized by looking at the TLFP values. The TLFP values from the three spectra from the first recording in rat 1(Fig. 3) were 128, 179, and 337%2. For the tumor BF signal measured by the second LDF probe, the TLFP values were 311, 307, and 219%2. The long-term TLFP values for all 21 measured tumor sites are summarized in Table 2.
Although most power spectra of the tumor BF recordings were dominated by the low-frequency fluctuations, there were also higher frequency components corresponding to respiration rate and heart rate, typically ∼50–60 and 300–400 cycles/min, respectively. These data are not shown here, because these higher frequency signals were not seen in the recordings and had no effect on fluctuations in oxygenation. Thus the fluctuations of interest in the current study are primarily those below 2 cycles/min.
Muscle was measured at seven sites in seven rats. The duration of the measurement period ranged from 77 to 92 min with a mean ± SD of 89 ± 5 min. In contrast with the tumors, the muscles were generally well oxygenated, and the electrode typically needed to be simply inserted into the muscle to find a location at which a nonzero could be recorded. The average during the first 10 min of recording was 14.0 ± 8.4 mmHg (means ± SD;n = 7) with a range of 5.1–23.5 mmHg and a median of 13.3 mmHg. Qualitatively muscle showed some fluctuations, but the fluctuations seemed to be much smaller in magnitude than those seen in the tumor. The tended to decrease over the course of the recording period, but it never dropped to 0 mmHg, as occurred in some tumors. At the end of the recording period, the average over the last 10 min was 10.9 ± 5.7 mmHg (means ± SD;n = 7) with a range of 4.0–17.4 mmHg and a median of 13.2 mmHg. The difference between the starting and ending was not significant (P > 0.05, Wilcoxon test).
An example of a muscle recording (rat 19) is presented in Fig.4 A. This particular example was chosen to show that a slow gradual drop in could occur in the muscle. Over the first 30 min of the recording, the decreased from ∼25 mmHg to 17 mmHg. For the next hour the was relatively stable, i.e., ∼16–17 mmHg with very small fluctuations.
Fourier analysis of muscle fluctuations.
The behavior described above is verified when the Fourier analyses for this experiment are examined. The short-term (5.5 min) analysis shows that there was only significant fluctuation in in the first few minutes of the recording and between 30 and 50 min (Fig.4 B). The average short-term TLFP value for this recording was 0.60 ± 0.59 mmHg2 (means ± SD,n = 32) with a range of 0.22–2.84 mmHg2 and a median of 0.32 mmHg2. The 6-min TLFP values for all of the muscle recordings are summarized in Table 1.
The 33-min spectra showed little additional information (Fig.4 C). Significant power was only seen in the first 33-min recording period, and this was dominated by the drift downward at the beginning of the period, which was manifested as a high-power peak at the lowest frequency. There was very little power at later times or at other frequencies. The long-term TLFP values for the four recording periods analyzed in this muscle were 8.79, 1.13, 0.79, and 0.67 mmHg2. The latter three values were typical for all of the muscle recordings except one, which showed a transient decrease in from 20 to 5 mmHg and had two TLFP values of near 50 mmHg2. The TLFP values for the 33-min power spectra from the muscle recordings are summarized in Table 1.
Muscle Blood Flow Fluctuations
Muscle BF recordings.
In addition to , muscle BF was simultaneously measured in the same 7 rats at 13 different sites. Similar to the tumor LDF measurements, all of the probes showed some fluctuation in muscle BF. The muscle BF traces from two LDF probes in the same muscle as described in the previous section are shown in Fig.5 A. The two traces of muscle BF were recorded over the same 90 min as the in Fig.4 A. Both probes showed relatively high-frequency changes in muscle BF with magnitudes of ∼5–10%. Whereas one probe remained relatively stable around the mean BF value (LDF 2), the other LDF probe fluctuated at a much lower frequency (LDF 1).
The short-term analysis of one of these muscle BF traces confirms these qualitative assessments (LDF 1; Fig.5 B). The signal fromprobe 1 showed a large peak in power at very low frequencies early in the measurement period. The magnitude of the peak was over 50%2, indicative of a change in amplitude of over 7% or a minimum-maximum change of over 14%. The power decreased over time during the experiment, and for the most part only smaller changes were evident at later times. The power spectrum for probe 2 was very flat, with all of the power below 6%2 and distributed almost evenly across the frequency range (not shown). Interestingly, the TLFP values for these two traces were similar. The average short-term TLFP value for the tumor BF recording from probe 1 was 32.8 ± 23.2%2 (means ± SD,n = 25) with a range of 12.2–108.1%2 and a median of 21.2%2. The mean fromprobe 2 was 32.5 ± 10.4%2 (means ± SD,n = 25) with a range of 18.7–53.6%2 and a median of 30.9%2. The 6-min TLFP values for all of the muscle BF recordings are summarized in Table 2.
The long-term analysis of the tumor BF recording fromprobe 1 (LDF 1) is shown in Fig.5 C. In this case, relatively little additional information is gained from the analysis. The first measurement period for probe 1included the large drop in muscle BF for that probe between 8 and 30 min (Fig. 5 A). This drop may have been responsible for the large decrease during the same time period (Fig. 4). The long-term analysis for probe 2 showed small-magnitude, low-frequency fluctuations for the earliest two measurement periods and very low-magnitude power during the last two 34-min periods (not shown). The long-term TLFP values for the 4 overlapping 34-min recording periods for the tumor BF recording fromprobe 1 were 239.8, 43.2, 25.0, and 42.9%2. The corresponding values for probe 2 were 47.0, 52.3, 37.9, and 35.3%2. The 34-min TLFP values for all of the muscle BF recordings are summarized in Table 2.
As was the case with the tumor BF analysis, most power spectra of the muscle BF recordings were dominated by the low-frequency fluctuations, but there were also higher frequency components present. Again these corresponded to respiration rate and heart rate, typically ∼50–60 and 300–400 cycles/min, respectively.
Comparison of Tumor and Muscle
Tumor and muscle fluctuations.
A comparison of the power spectra from muscle and tumor revealed no difference between the peak frequencies (Fig.6 A). None of the first 20 peak frequencies was statistically different between the two tissues (Mann-WhitneyU-test,P > 0.12).
From the examples already shown, the fluctuations in tumor appeared to be of a higher magnitude than those seen in muscle. This pattern was true for all of the experiments in this study. The magnitude of the changes in the two tissue types is most easily compared by examination of the TLFP from 0 to 10 cycles/min (Table 1). The most meaningful comparison is between the long-term TLFP values. Only 6 of the 27 muscle recording periods (22%) had a TLFP above 3 mmHg2, whereas 32 of the 48 tumor sites (67%) showed TLFP values above that level. The tumor sites had significantly higher TLFP than the muscle (P < 0.001; Mann-WhitneyU-test). If the median TLFP values for every measurement site are compared, the tumor sites had significantly higher TLFP than the muscle (P = 0.02, Mann-Whitney U-test). Only 1 of the 7 muscle recording sites (14%) had a median TLFP above 2 mmHg2, whereas 11 of the 13 sites in tumor (85%) had median TLFP values above that level.
Another type of comparison can be made by examining the cumulative power of the power spectra for each tissue type over a given frequency range. This comparison is made in Fig. 7for the 33-min period data. In Fig. 7 as well, it is evident that fluctuations in tumor were larger than those observed in muscle. This was true not only for the slowest fluctuations (0.03 cycles/min), but also at the other low frequencies as well. This is manifested by the steeper slope of the curve for the tumor as frequency increases.
Finally, the magnitude of the fluctuations in the two tissues can be compared more rigorously by looking at the power at each peak frequency. Figure6 B compares the power at the first 10 frequency peaks of the 33-min power spectra in muscle and tumor. There is clearly increased power in tumor at every frequency, even at the higher frequencies where the power is relatively low.
Tumor and muscle BF fluctuations.
A direct comparison of the peak frequencies in muscle BF and tumor BF showed that there was more low-frequency BF activity in the tumor (Fig.8 A). For example, the median sixth peak frequency in tumor was only 0.468 cycles/min, compared with 0.558 cycles/min in muscle (P < 0.001; Mann-WhitneyU-test).
The examples of BF already shown for both tissues demonstrate that significant low-frequency fluctuations in flow can occur in tumor and in muscle. As noted earlier, there can also be differences in the magnitude of fluctuations within the same tumor as measured by different LDF probes, indicating local intratumoral variation of tumor BF. However, from the individual examples presented so far, it is not apparent if any differences in BF fluctuations exist between muscle and tumor.
The median TLFP values from the power spectra of tumor and muscle LDF BF are compared in Table 2. Again, the long-term TLFP values are probably more meaningful, because they yield higher frequency resolution in the region of interest (0–10 cycles/min). There was no difference between the tumor TLFP values and the muscle TLFP values when the values for all of the LDF probes at all tissue measurement sites were compared (P = 0.107, Mann-Whitney U-test). However, if all of the 33-min periods were considered separately, then the TLFP of tumor was greater than muscle (P = 0.01, Mann-Whitney U-test). This latter comparison is more representative of the true data, because averaging multiple spectra at a specific measurement site tends to mask fluctuations, which have been shown to differ during different 33-min analysis periods in the same site (Figs.3 C and5 C).
The cumulative power of the LDF power spectra for each tissue type is displayed as a function of frequency for the 33-min period data in Fig.9. In Fig. 9, it is also evident that fluctuations in tumor BF were different from those observed in muscle BF. At frequencies from 0 to 7 cycles/min, the slope of the tumor BF curve is steeper than that of the muscle BF curve, indicating that there was much more power being contributed at these frequencies in tumors. This supports the finding that there was more activity at low frequencies in tumors compared with muscles (Fig.8 A). This conclusion is further substantiated by looking at the peak powers at each peak frequency from the muscle BF and tumor BF power spectra (Fig.8 B). At almost all of the frequency peaks in the low-frequency range (between 0.2 and 2.0 cycles/min), the power in the tumor BF signal was greater than that in the muscle BF recording (P < 0.02, Mann-WhitneyU-test).
Possible Correlation Between Blood Flow and
Tumor BF and fluctuations: frequency and magnitude.
The relationship between tumor and tumor BF is summarized for all of the data in Fig.10 A. The peak frequency comparison showed a clear upward shift above the identity line. Beyond the fourth power peak, the frequencies of the power peaks were higher than those of the corresponding peaks in the tumor BF power spectra (P < 0.006,n = 79 spectra). In other words, there tended to be more power peaks in the tumor BF signal compared with the signal, indicating slightly more low-frequency activity in the tumor BF than in the signal.
This finding was substantiated when power spectra peak frequencies from the individual experiments were correlated and compared. In Table3 the correlations between tumor peak frequencies and the peak frequencies of the tumor BF fluctuations are summarized. The values can be interpreted in the following manner. Shallower slopes (<1.0) would be associated with the points lying below the identity line, because the first frequency is almost always 0.03 cycles/min. Conversely, slopes above 1 are indicative of lines above the identity line. For the tumor and tumor BF experiments, the median slope was 1.066 with over 73% of the values greater than 1. The 75th percentile value was also well above 1, indicating that most of the individual relationship lines were above the identity line. Thus there tended to be more low-frequency activity in BF than in .
Muscle BF and fluctuations: frequency and magnitude.
The relationship between BF and in the muscle was different from that seen in the tumor (Fig.10 B). In the muscle, the peak frequencies of the power spectra and the muscle BF spectra were almost identical, indicating that there was some activity at almost all of the same frequencies in both signals (P > 0.25,n = 50 spectra). When the power spectra peak frequencies from the individual experiments were compared, the slopes of the correlations were spread out almost equally below and above 1 with 46% of the slopes under 1 (Table 3).
Possible Role of BP Fluctuations
All animals in the study had MABPs in the normal range for rats (42). In the tumor-bearing animals, the average MABP during the first minute of recording was 102.4 ± 9.4 mmHg (mean ± SD;n = 13) with a range of 86.0–122.7 mmHg and a median of 103.5 mmHg. By the end of the experiment, the MABP was 99.7 ± 11.2 mmHg with a range of 85.2–113.7 mmHg and a median of 99.8 mmHg. The rats in which muscle and BF were measured had similar MABP. At the beginning of the recording time, the MABP was 102.8 ± 12.1 mmHg (mean ± SD;n = 7) with a range of 92.2–125.6 mmHg and a median of 97.6 mmHg. By the end of the recording period, the MABP was 105.4 ± 6.3 mmHg (mean ± SD;n = 7) with a range of 94.1–111.7 mmHg and a median of 106.1 mmHg.
The power spectra of the blood pressure recordings were dominated by the higher frequency components corresponding to respiration rate and heart rate, typically ∼50–60 and 300–400 cycles/min, respectively. Usually there was also some low-frequency power, in the range of 1–10 cycles/min. This power was evident in many tumor and muscle experiments, in which BP could fluctuate by 5–10 mmHg during the recording time (data not shown).
Tumor and muscle and BP fluctuations: frequency and magnitude.
The major reason for examining the BP traces using Fourier analysis was to determine if changes in BP were related to changes in tissue or BF. To test whether there was any relationship between tissue and BP, the frequencies of the power peaks in the two power spectra were compared, as was done for and BF. An examination of power spectra from the individual experiments revealed that there were substantial differences between the power spectra in tumors (Table 3). The median slope was 0.998 with exactly equal spread in the values above and below 1. The 25th and 75th percentile values were also very far from 1, indicating large spread in the data. Thus there was no clear relationship between tumor and BP.
In the muscle there appeared to be more very low-frequency activity in the trace than was evident in the BP recordings, but there was clearly no consistent pattern in the correlations (Table 3). The median slope was 0.906, indicating that there was a tendency for the peak frequencies to be lower than the corresponding BP frequencies. Despite the shallower median slope, this trend was not significant, because the spread as indicated by the 25th and 75th percentiles was very large.
Tumor and muscle BF and BP fluctuations: frequency and magnitude.
The relationship between tissue BF fluctuations and BP fluctuations was also examined by comparing the frequencies of the power peaks. There was more low-frequency activity in the tumor BF spectrum than in the BP spectrum, as evidenced by a significant shift of the correlation lines to the right of the identity line. The median slope of the 79 correlation lines was 0.889, and 77.2% of the correlations had slopes under 1.
Similarly, there was a tendency for the muscle BF to have more activity at frequencies ∼1 cycle/min than shown in the BP spectrum. The median slopes for all 50 regressions was 0.967 with 25th and 75th percentiles of 0.860 and 1.027, respectively, and 64.0% of all the lines had slopes under 1 (Table 3). Thus there was a tendency for the lines to be below the identity line, indicating more low-frequency activity in the BF signals compared with the blood pressure recordings. However, this tendency was less pronounced in muscle than that seen in tumors.
In this study temporal changes in tumor and muscle oxygenation have been characterized for the first time using Fourier spectral analysis. in both tissues fluctuated at very low frequencies (<1–2 cycles/min), but the magnitudes of fluctuations in tumor were larger than those seen in muscle. Laser-Doppler BF signals were simultaneously measured in the tissues and were also subjected to Fourier analysis. Both tissues showed low-frequency fluctuations in BF, but there was more low-frequency activity in the tumor and the magnitude of the BF fluctuations tended to be higher in tumor than in muscle. In tumor, there was more low-frequency activity in BF compared with , whereas in muscle there was no significant difference between the peak frequencies of BF and . None of the fluctuations could be directly correlated with changes in BP.
Decrease in Mean Tumor During Measurement Period
In this study, tumor decreased from 15.2 ± 10.5 mmHg (mean ± SD;n = 13) at the beginning of the measurement period to 6.9 ± 9.3 mmHg at the end (P < 0.01; Wilcoxon test). The reason for the significant decrease in tumor over the course of the experiment is not clear, but it is most likely related to the selection process used in determining a suitable recording location, rather than a measurement artifact of the electrodes. The oxygen microelectrodes are specifically designed to minimize tissue damage and are ∼10 μm in diameter at the tip. Significant tissue damage due to the electrodes would most likely result in a much faster drop in than the slow decreases seen here. The decrease in was also not a result of oxygen consumption by the electrode, because these recessed electrodes are designed to minimize consumption and the resultant disturbance of the diffusion field (39). In addition, there are numerous examples where there were decreases followed by increases in , a behavior inconsistent with a drop in caused by consumption of oxygen in a closed system. Most likely this decrease is real and is a consequence of our initial selection process of a nonzero . This is more evident when the 13 experiments are examined individually.
The behavior of the tumor can be roughly subdivided into three different types. In four cases the tumor gradually dropped to zero after 22–68 min of recording and remained at zero for the duration (∼20 to 40 min). In three other cases, the fluctuated up and down during most of the experiment but eventually ended up at a mean much lower than the starting value. In one example, the started at 29 mmHg, rose to 40 mmHg after 55 min, and decreased to 12 mmHg by 90 min. In the other six cases, the was fluctuating around a new value within 6 mmHg of the starting value. In two of these six experiments the end value was higher. An example in which the ending was lower is shown in Fig.1 A. Whereas the four cases in which dropped to zero may have possibly been caused by some local tissue damage, we could not prove this and saw no reason to exclude these data. As already noted, we selected a starting that was clearly distinguishable from zero to permit the to transiently fluctuate up or down. A point already at zero could easily have been located in a chronically hypoxic area, where the would have remained at zero throughout the experiment, regardless of local changes in flow. By choosing a point with a relatively high , we were also selecting for points which would more likely decrease than increase. This is particularly true for the cases that dropped more than 6 mmHg, including those which dropped to zero. If the was high initially because it was near a blood vessel, a decrease in local perfusion would have dropped the significantly, even to zero. The converse of that scenario is much less likely, because we would never have selected a point near a closed or low-perfused vessel that might have increased its flow later. Therefore, the selection was biased toward values that might decrease by large amounts but would most likely not increase by similar magnitudes.
Tumor and Muscle Fluctuations
There have only been a few other studies of fluctuations in tissue, and most of those were in neural tissue. Lübbers and Leniger-Follert (29) observed changes in in the brain at frequencies near 1 cycle/min. Manil et al. (30) performed the first quantitative spectral analysis of fluctuations. They measured in the cerebral cortex of awake rabbits using chronically implanted electrodes and found that there were significant fluctuations in the brain at low frequencies. The power gradually decreased over the range of 0 to ∼24 cycles/min, with peaks superimposed at 3.6, 10.2, and 21.6 cycles/min. Riva and co-workers (36) measured BF and in the cat optic nerve head and found that fluctuated at frequencies ranging from 2.5 to 4.5 cycles/min, in phase with the BF parameters. Another study (1) investigated spontaneous fluctuations in as a function of position in the retina of cat and monkey. In portions of the vascularized retina, power spectra similar to those found by Manil and co-workers (30) were obtained. Significant power was distributed between 0 and 30 cycles/min and sometimes beyond. The only observations (47) made on transient changes in in a nonneural tissue were those from the gracilis muscle of guinea pig, in which fluctuated irregularly at ∼1–2 cycles/min.
The fluctuations in tumor and muscle reported here occurred over a much lower frequency range than those reported in neural tissue, but the range is very similar to that noted for gracilis muscle. This is not too surprising, because neural tissue would be expected to have very fine control over BF and oxygenation. Because muscle can and often does carry out glycolysis, it is not essential for that tissue to maintain finely controlled oxygen levels. Tumor, of course, would be expected to exert little or no control over its oxygen supply. Alternatively, brain and retina might require more exquisite, relatively high-frequency control of oxygen levels. This difference in the needs and natures of the tissues may explain the difference in the characteristics of the fluctuations found in the tissues.
The fluctuations in must be caused by either changes in oxygen delivery or oxygen consumption or by a combination of the two. Although changes in oxygen consumption cannot be ruled out, it is difficult to derive a hypothesis to explain fluctuations based solely on changes in oxygen consumption. Although not totally dismissing the possibility of changes in consumption, most previous studies have attributed the changes in local primarily to alterations in oxygen supply, i.e., BF (1, 4, 47). If the fluctuations are indeed primarily attributable to changes in local BF, it is necessary to postulate circulatory or hemodynamic mechanisms that could be responsible for the changes in at the observed frequencies. Possible mechanisms behind the and BF fluctuations will be discussed in a later section.
Tumor and Muscle BF Fluctuations
LDF has been used previously to measure BF fluctuations in various tissues, including muscle (27, 37, 38) and experimental and human tumors (6, 18). These fluctuations in LDF are typically termed “flux motion” or “flow motion” and have been analyzed in several different ways, including Fourier analysis (8) and frequency histogram techniques (21). In most tissues the flow motion has been found to occur at frequencies between 1 and 10 cycles/min. This high-amplitude, low-frequency flow motion is most likely a result of arteriolar activity or vasomotion (8). The rhythmic diameter changes associated with vasomotion typically occur in vessels below 100 μm in diameter and at frequencies ranging from 1 to 15 cycles/min (7). Sometimes low-amplitude, high-frequency fluctuations (∼20 cycles/min) are also present, and these are most likely caused by respiration-induced changes in venular BF (8).
In this study almost all of the spontaneous fluctuations in BF occurred at very low frequencies below 2 cycles/min. Usually very little activity of significant magnitude was seen between 2 and 30 cycles/min. In most power spectra there was evidence of activity at higher frequencies corresponding to respiration rate and heart rate, typically ∼50–60 and 300–400 cycles/min, respectively. This was true not only of the tumor tissue, but of the muscle tissue as well. Thus when these data are compared with previous LDF data, several differences must be kept in mind. First, unlike the results from other studies in other tissues, there was little activity between 2 and 30 cycles/min. This means that there was an absence of activity in the BF signal at those frequencies typically associated with vasoactivity. Second, the significant low-frequency activity was below 1–2 cycles/min, where most other investigators found little or no activity. The reasons behind these discrepancies could be real or artifact.
There are several possible reasons for a lack of significant LDF activity at frequencies associated with vasomotor activity. One possibility may have been the use of pentobarbital as the anesthetic. It has been shown that intravenous doses of 35 mg/kg can cause loss of vasomotor activity in hamster vessels (7). Whereas it may be possible that pentobarbital diminished vasoactivity in the present study, it is unlikely that it would have totally blocked any spontaneous activity, because we have previously demonstrated diameter fluctuations in the window chamber preparation in the same Fischer 344 rats under the same pentobarbital anesthesia under control conditions (12) and during hypotensive episodes (41). Another possibility is that these spectra represent the actual BF fluctuations in tumor and muscle and that these tissues simply do not exhibit higher frequency vasoactivity (2–20 cycles/min). Certainly this could well be true in tumor, where there is no direct evidence of high-frequency fluctuations in BF, but there is evidence of very low-frequency (<0.5 cycles/min) fluctuations in both murine and human tumor LDF (6, 18). In addition very slow fluctuations in erythrocyte flux have been demonstrated in microvessels of a transplanted tumor in the rat (26). Interestingly, the humans in whom slow LDF fluctuations were found were not anesthetized, which again suggests that anesthesia has little effect on this type of fluctuant behavior. One might expect muscle to show more classical vasoactivity over the entire frequency range; however, there are several examples of LDF recordings in nontumor tissues, including muscle, which showed little activity at frequencies between 10 and 30 cycles/min. For example, Buerk and Riva (5) showed that most of the activity in the power spectra of LDF recordings in the cat optic nerve head was below 10 cycles/min. The vast majority of the fluctuations were between 0 and 5 cycles/min. In skeletal muscle, LDF recordings under control conditions did not show fluctuations at frequencies associated with vasomotion (27).
There is little or no mention in the literature of slower fluctuations at frequencies under 1 cycle/min, although some studies show significant power peaks at 1 or 2 cycles/min (5, 21). This is most likely a result of the short durations of the measurement periods analyzed in previous studies. Typically, BF is measured for 5 min or less (e.g., Refs. 5 and 21). Because the frequency resolution of the spectrum is a function of the number of sampled points, the shorter the measurement period is, the lower the resolution will be. For example, in this study, 5.5-min records were analyzed using Fourier analysis (Figs. 1 B,2 B, and5 B). The resolution of the power spectra from these recordings was 0.00305 Hz or 0.1831 cycles/min. This translates into five points between 0 and 1 cycles/min. To improve the very low-frequency resolution, we chose to analyze recordings of over 33 min, which yielded spectra with a resolution of 0.03 cycles/min. Such long recordings of laser-Doppler BF have not been analyzed before. Thus in this study high-resolution analysis of the low-frequency activity in LDF signals was performed for the first time.
Laser-Doppler measurements of BF have previously been performed in experimental and human tumors (6, 18). The overall findings were that BF fluctuates greatly in tumors at frequencies of several cycles per hour and that the changes are heterogeneous, even within the same tumor. We have previously noted this instability in our own laboratory using the Oxford Doppler Array system, and this is demonstrated again in the present study (Fig. 3 A). In that example, BF could fluctuate by as much as 40–50% of the average value. Such results are very consistent with other LDF measurements made in tumors (6, 18).
The muscle BF also fluctuated at very low frequencies in these studies. Kvernebo and co-workers (27) measured LDF in human skeletal muscle and made no mention of vasoactive fluctuations in the signals. The examples shown in that study give no evidence of flow motion occurring under control conditions. In a series of studies examining LDF in rabbit skeletal muscle, Schmidt and co-workers (37, 38) demonstrated no slow wave flow motion (1–3 cycles/min) under normal conditions, but occasionally found some higher frequency fluctuations. Thus the lack of typical flow motion activity in the muscle LDF recordings in the present study are in overall agreement with previously published results.
Possible Mechanisms of BF and Fluctuations
There was some evidence of a direct correlation between the muscle fluctuations and muscle BF fluctuations, but there was significantly more low-frequency activity in tumor BF compared with tumor (Fig. 10). This suggests that the muscle BF fluctuations may cause at least some of the fluctuations in in that tissue. In the case of the tumor, the link between and BF may not be as clear. This might be expected because some of the blood in tumor vessels is highly deoxygenated (12, 26), and a change in BF may not necessarily translate into a change in oxygen delivery in all areas of the tumor. Of course, it is also important to remember that for both tumor and muscle BF and were measured in different locations. The laser-Doppler probes would also sample a much larger volume than a 10- to 20-μm-diameter oxygen microelectrode and would average the BF over the volume (43). Thus the two probes in this study were measuring local parameters at different locations with different sampling volumes. Nevertheless, it seems most likely that the fluctuations in are some reflection of changes in local perfusion of the measurement area. Therefore, some possible mechanisms to produce low-frequency fluctuations in BF will be explored in an effort to suggest possible mechanisms behind the observed fluctuations in tumor and muscle.
The most commonly forwarded explanation for changes in local BF is vasomotion. The rhythmic diameter changes associated with vasomotion typically occur in arterioles below 100 μm in diameter and at frequencies ranging from 1 to 15 cycles/min (7), although this frequency range may be tissue dependent. For example, Hundley and co-workers (22) demonstrated much lower frequencies of vasomotion in cerebral arterioles in awake rabbits. They found the mean frequency of diameter changes to be 0.74 cycles/min, with a range of 0.4–3.5 cycles/min. The nature of vasomotion also changes with vascular diameter. Using a hamster skinfold window preparation, Colantuoni and co-workers (7) found a negative correlation between the fundamental vasomotor frequency and arteriolar diameter and between relative vasomotor amplitude and diameter. Small arteries, with diameters of 70–100 μm, usually fluctuated by 8–30 μm (19% mean amplitude change) at frequencies of 1–3.5 cycles/min (mean of 2.7 cycles/min). Alternatively, terminal arterioles (6- to 15-μm diameter) typically showed rhythmic activity at 6–15 cycles/min (mean of 9.5 cycles/min) with diameter changes of 2–11 μm (89% mean amplitude change).
It appears unlikely that vasomotion can explain all of the fluctuations seen in the present study, because typical vasomotor activity is usually higher than what was observed here. Of course, some vessels have been noted to fluctuate in diameter at frequencies below 1 cycle/min (22). In tumors growing in the rat window chamber model, we have noted fluctuations in tumor-feeding arteriolar diameter at frequencies of 0.6 cycles/min under hypotensive conditions (41). Thus there is some evidence that some arterioles, particularly in tumors, are capable of these very slow fluctuations in diameter. Nevertheless, not all of the very slow fluctuations (<0.5 cycles/min) seen in the present study can be necessarily attributed to vasomotion. If vasoactivity is accountable for at least some of the observed fluctuations in BF and , then it would have to be attributed to the larger arterioles (70–100 μm) because these vessels typically demonstrate classical low-frequency vasomotion at ∼1–2 cycles/min (7).
BF changes via hemodynamic mechanisms.
Another possible explanation for the very low-frequency fluctuations seen in this study was put forward by Kiani et al. (25). In a very interesting study (25), they attempted to show the impact of hemodynamic properties on erythrocyte velocity in the microvasculature. They developed a model of network BF based on experimental measurements made in the rat mesentery, and they compared the model results to in vivo measurements of erythrocyte velocity and discharge hematocrit. The model accounted for several hemodynamic characteristics, including nonlinear flow properties of blood, unequal distribution of erythrocytes at vessel bifurcations, and nonuniform axial distribution of erythrocytes within vessels (25). Several significant results came out of the study. First, the model generated temporal variation in erythrocyte velocity and discharge hematocrit with fixed vessel dimensions and constant network flow, indicating that the fluctuations were due solely to hemodynamic factors. Second, vessels arising from the same parent vessel node could demonstrate widely different fluctuation patterns. Third, a Fourier analysis of arteriolar (11- to 48-μm diameter) erythrocyte velocity revealed that the maximal peak frequencies were in the range of 0.05–0.5 Hz (3–30 cycles/min). These patterns of oscillation were similar to those predicted by the model. The observed fluctuations were not caused by vasomotion, because the authors (25) stated that vasomotion was rarely seen in their preparations and they also found these fluctuations when the vessels were maximally dilated.
Again the reported frequencies of Kiani et al. are slightly higher than what was found in the current study. Nevertheless, the BF fluctuations in tumor and muscle may be attributable to this phenomenon. In the study of Kiani and co-workers (25), it is stated that erythrocyte velocities were recorded at 30 Hz for 3 min, and these recordings were subjected to Fourier analysis. However, in the example shown in Figs. 4and 5 of Ref. 25, only a 30-s recording was analyzed (25). Thus in that example 900 points were available for the analysis, which translates into a frequency resolution of 0.0333 Hz or 2.0 cycles/min. This is much lower than the frequency resolution obtained in even the short-term Fourier analysis of the LDF signals in the present study (0.144 cycles/min; Figs. 3 B and5 B). Thus there may have been undetected significant power below 2 cycles/min in their erythrocyte velocity data as well.
In addition to the hemodynamic factors considered by Kiani et al. (25), there are other hemodynamic properties that could have an effect on BF, particularly in tumors. Tumors have many properties that would make them very prone to fluctuations in BF. For example, they are known to be very permeable, and plasma can freely cross the vessel wall (48). Any changes in the fluid exchange between the vessel and tumor interstitium would result in a change in microvascular hematocrit and affect blood viscosity. Such alterations in fluid exchange could occur at a frequency of ∼1 cycle/min and could be partially responsible for the BF fluctuations observed in this study. In addition, tumors have a very chaotic, tortuous vasculature, which would most likely result in unstable intravascular pressures. Tumor vessels have even been shown to be transiently hypoxic (26), and this could result in a transiently increased RBC viscosity and fluctuations in BF. Whether any or all of these hemodynamic mechanisms occur in tumors and are responsible for BF fluctuations remains to be determined.
Intussusceptive microvascular growth.
Another possible mechanism for the fluctuations in tumor is rapid vascular remodeling via intussusceptive microvascular growth (34). Intussusceptive microvascular growth is a process of vascular network formation characterized by insertion of interstitial tissue columns into the vessel lumen with subsequent partitioning of the lumen. In a tumor window preparation in rat, it was shown (34) that formation of the tissue pillars occurred at a frequency ranging from 0.8 to 2.3 pillars/h, depending on the age of the tumor. Patan and co-authors (34) suggested that the architectural changes resulting from intussusceptive microvascular growth might be responsible for intermittent tumor BF.
Most likely, intussusceptive microvascular growth does not account for all of the fluctuations in BF and observed in the present study, because similar frequency patterns were seen in both tumor and muscle. Even though intussusceptive growth has been demonstrated in normal, uninjured adult rat skeletal muscle (16), one would not typically expect it to occur in muscle to the extent that it does in tumor. Also, the frequencies associated with intussusceptive microvascular growth (0.8–2.3 events/h or 0.013–0.038 cycle/min) are all slightly lower than some of the observed fluctuations in BF and in the current study. Nevertheless, such changes within the microvasculature could interact with vasomotor or hemodynamic factors to affect local BF and local over the frequency range seen in this study.
Evaluation of possible mechanisms behind fluctuations.
Based on the previous discussion of possible mechanisms underlying BF and fluctuations, it seems most likely that all three may play some role in the changes seen in the current study. Classical vasomotion of larger arterioles may result in oscillations of BF at 1–2 cycles/min, and this would be translated into fluctuations of similar frequencies. The very low-frequency fluctuations in flow and may well be caused by a combination of intussusceptive microvascular growth and hemodynamic mechanisms. Of the three mechanisms, the hemodynamic effects may be the most significant in tumors. Tumors are characterized by chaotic vascular networks consisting of tortuous, irregular vessels. In addition, the blood within the vessels can be very hypoxic, leading to changes in the mechanical properties of blood, including an increase in erythrocyte viscosity (23). Therefore, tumor blood vessels are probably much more likely to exhibit the hemodynamic changes modeled by Kiani and co-workers (25), which resulted in low-frequency BF fluctuations.
Differences in Fluctuations in Tumor and Muscle
Although tumor and muscle demonstrated similar low-frequency fluctuations in both BF and oxygenation, there were several key differences between the oscillations in the two tissues. Although it is true that BF fluctuated over a similar low-frequency range in both tissues, there was significantly more low-frequency activity in the tumor. In addition, the magnitude of the BF oscillations in the tumor tended to be greater in the tumor compared with the muscle. The fluctuations occurred over similar frequency ranges in tumor and muscle, but the magnitude of the changes was significantly greater in the tumor. Thus, in general, tumors demonstrated more large amplitude, low-frequency fluctuations in BF and compared with muscle. To appreciate the importance of these differences, it is necessary to understand the impact of low-frequency fluctuations in BF on tissue oxygenation.
Effects of BF fluctuations on tissue .
There have been at least two significant studies concerning the effects of changes in BF on tissue oxygenation. In the first, Secomb and co-authors (40) used theoretical models to predict the effects of vasomotion on oxygen delivery. They found that only fluctuations in BF with lower frequencies produced appreciable fluctuations in at significant distances away from the vessel. In other words, the penetration distance of oxygen increased as the fundamental frequency of the vasomotion decreased. The low-frequency fluctuations associated with vasomotion (in their example 0.2 Hz or 12 cycles/min) were capable of providing intermittent oxygenation to areas that would normally be prone to hypoxia or anoxia (40). In a later study with a different model, Tsai and Intaglietta (45) verified the primary finding of Secomb et al. They also showed that lower frequencies of flow motion resulted in an increase in the volume of tissue that achieved a minimum level (e.g., 5 mmHg). In addition, they noted that high amplitudes of flow motion resulted in a decrease in average tissue , an increase in the variability of tissue , and an increase in the variability of the amount of oxygen delivered by each erythrocyte (45).
Significance of BF fluctuations on tumor .
In the present study, the fact that tumor tissue consistently showed more low-frequency, high-amplitude oscillations in BF compared with muscle is of major importance. Because low-frequency fluctuations in BF result in greater penetration of oxygen into the tissue and permit oxygenation of a larger volume of tissue, the presence of significant low-frequency fluctuations in tumor BF would mean that diffusion distances would be greater in a tumor experiencing fluctuations in BF than if it were experiencing steady BF. This would be of critical importance to a tumor with low vascular density and large intervascular distances, but it would be of less importance to a normal, homogeneously vascularized tissue like muscle. In this study, it was also shown that the BF and fluctuations in tumor were larger in magnitude than in the muscle. High amplitudes of flow motion theoretically result in a decrease in average tissue and a greater variability in local tissue (45). The presence of these large fluctuations would produce greater heterogeneity in the tissue oxygenation profile than would be seen with smaller fluctuations. This characteristic, along with the tortuous path length of tumor vessels arising from limited arteriolar inputs (13), results in a tissue that has very unstable BF, unstable and heterogeneous oxygenation, and areas of transient hypoxia.
Clinical Significance of BF and Fluctuations in Tumor
The presence of hypoxia in tumors is known to play significant roles in tumor response to radiation therapy (44), selection for more aggressive phenotypes (15), mutation frequency (35), and treatment outcome and patient survival (3, 19, 20, 32, 33). Thus hypoxia is an important physiological parameter in tumors, and an understanding of how and why it occurs is crucial to an understanding of tumor physiology and tumor treatment response.
Unfortunately, hypoxia in tumors may be much more complicated than previously thought. It is typically stated that two types of hypoxia exist: chronic or diffusion-limited hypoxia and acute or perfusion-limited hypoxia. Chronic hypoxia supposedly develops because of unusually long diffusion distances between tumor vessels (44). Acute hypoxia has classically been thought to develop because of collapse of or transient blockages in tumor vessels (33).
Only recently has more information arisen to demonstrate that these two types of tumor hypoxia are much too complex to be defined so simply. For example, it is now clear that distances between tumor vessels or vascular density may not be the only major determinant of whether the tissue between them is hypoxic. Many vessels in tumors carry severely deoxygenated blood, so their ability to supply oxygen to the tumor is limited (17, 26). These well-perfused but hypoxic vessels result in hypoxic areas in the neighboring tumor interstitium (17). Because the volume of tissue supplied with oxygen by a particular vessel is critically dependent on the vascular and the consumption rate of the intervening cells, the intervascular distance is often not the oxygen-limiting factor for the tumor parenchyma between the vessels. Thus chronic hypoxia cannot be simply defined as hypoxia that develops because of the distance of the cells from the surrounding vessels. Similarly acute hypoxia is also difficult to succinctly define. For example, it is now clear that tumor hypoxia can result without full closure of vessels (26). Perivascular has been shown to be well correlated with the erythrocyte flux in tumor vessels, and the perivascular can often transiently drop below 5 and 10 mmHg (26). This would clearly result in a fluctuation of tumor parenchymal , and some cells could oscillate in and out of radiobiological hypoxia. Such basic physiological work in tumors has suggested that this transient hypoxia exists in solid tumors, but the time course and magnitude of the parenchymal fluctuations has not been definitively shown until now.
This study has clearly demonstrated that transient hypoxia exists in solid experimental tumors and that it can fluctuate at very low frequencies, which may affect treatment outcome. The recording shown in Fig.1 A is a dramatic demonstration of the presence of radiobiologically significant changes of in tumors. In this example, there are many periods of several minutes during which the is below 3–5 mmHg, which is considered to be the critical for radiosensitization. Obviously if radiotherapy were administered during that time, the cell killing in this portion of the tumor would be poorer than anticipated. Other examples of tumor fluctuating between normoxia and hypoxia are shown in a previous report (11).
It is interesting to speculate on the best ways to manipulate tumor to therapeutic advantage. As demonstrated in the present study, the large, low-frequency fluctuations in local BF most likely lead to the large tumor changes. High-amplitude, low-frequency fluctuations in BF lead to deeper penetration of the oxygen into the tumor, but also to a more heterogeneous distribution of and a lower overall average (45). Because vascular levels in tumor may be very low, this may result in many cells receiving the minimal amount of oxygen they need to survive. A more stable BF might result in a smaller viable concentric ring around the vessel with a surrounding annulus of essentially chronically hypoxic tissue. Whether this is advantageous to therapy is a difficult question to answer. For example, stable BF would probably increase the cell killing near the vessel during radiation therapy, but fewer cells further from the vessel would be destroyed. On the other hand, the administration of a hypoxic cytotoxin while flow was fluctuating, followed by an agent to inhibit the oscillations in flow, could result in a decrease in far from the vessel and an increase in cell killing. Obviously, this is a complex issue that requires further investigation.
The impact of fluctuations in BF and in tumors on various therapeutic modalities remains to be determined. Certainly their presence complicates an already complex physiological system, in which the effects of therapeutic manipulations are often difficult to accurately predict. Future work is necessary to determine the exact mechanisms behind these fluctuations and to attempt to modulate them. Finally, it needs to be determined if it would be advantageous to minimize or maximize these fluctuations during treatment to optimize therapeutic gain.
We thank Dr. Tim Secomb and Garheng Kong for input and helpful, thought-provoking discussions. We also thank Dr. David M. Brizel for a careful reading of the manuscript and for helpful suggestions.
Address for reprint requests: R. D. Braun, Dept. of Radiation Oncology, Box 3455, Duke Univ. Medical Center, Durham, NC 27710 (E-mail:).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. §1734 solely to indicate this fact.
- Copyright © 1999 the American Physiological Society