In1 J Rodmfon Oncology Bwl Ph'hp Vol. 18, pp. 5X-568 Pnnted I” the U.S.A. All rights reserved.

copyright

0360.3016/90 $3.00 + .oO (c‘ 1990 Pergamon Press plc

0 Original Contribution HETEROGENEITY

IN TUMOR

M. W. DEWHIRST, C. GUSTAFSON, ‘Duke University

Medical Center,

MICROVASCULAR

RESPONSE

TO RADIATION

D.V.M., PH.D.,’ R. OLIVER, B.S.,’ C. Y. Tso, PH.D.,’ B.S.,’ T. SECOMB, PH.D.~ AND J. F. GROSS, PH.D.~ Durham,

NC 27710, and 2Arizona Health Sciences Center, Tucson,

AZ 85724

Viable hypoxic cells have reduced radiosensitivity and could be a potential cause for treatment failure with radiotherapy. The process of reoxygenation, which may occur after radiation exposure, could increase the probability for control. However, incomplete or insufficient reoxygenation may still be a potential cause for local treatment failure. One mechanism that has been thought to be responsible for reoxygenation is an increase in vascular prominence after radiation. However, the effect is known to be heterogeneous. In this study, tumor microvascular hemodynamics and morphologies were studied using the R3230 AC mammary adenocarcinoma transplanted in a dorsal flap window chamber of the Fisher-344 rat. Measurements were made before and after (at 24 and 72 hr) 5Gy radiation exposure to assess microvascular changes and to explore possible explanations for the heterogeneity of the effect. There was considerable heterogeneity between tumors prior to radiation. Vascular densities ranged from 67 to 3000 vessels/mm3 and median vessel diameters from 22 to 85 pm. Pretreatment perfusion values varied by a factor of six. In irradiated tumors, conjoint increases in both vascular density and perfusion occurred in most tumors, although the degree of change was variable from one individual to the next. The degree of change in density was inversely related to median pretreatment diameter. Relative change in flow, as predicted by morphometric measurements, overestimated observed changes in flow measured hemodynamically. These results support that heterogeneity in tumor vascular effects from radiation are somewhat dependent on pretreatment morphology as well as relative change in morphology. Since changes in flow could not be completely explained by morphometric measurements, however, it is likely that radiation induced changes in pressure and/or viscosity contribute to the overall effect. Further work in this laboratory will investigate these hypotheses. Microcirculation,

Reoxygenation,

Hypoxia, Heterogeneity.

INTRODUCTION

mechanisms for the observed variability in tumor vascular responses to radiation. To accomplish this goal, detailed morphometric and hemodynamic measurements were made prior to and at 24 and 72 hr after radiation. The model system was the R3230 AC adenocarcinoma transplanted in a dorsal flap window chamber of the Fisher 344 rat. Using the Hagen-Poiseuille equation (18), morphometric measurements were used to predict relative flow changes. The predicted changes were compared with actual measured changes in flow. Discrepancies between the predicted and measured data provide clues regarding mechanisms of the effect of radiation on tumor microvasculature.

Viable hypoxic cells have reduced radiosensitivity which may lead to treatment failure from radiotherapy. The process of reoxygenation, which has been shown to occur after radiotherapy may partially alleviate their significance (9, 19, 20, 27, 28, 36). In spite of this process, however, it is possible that it is incomplete or inefficient in some tumors, thus leading to treatment failure.

There is reason to hypothesize that the reoxygenation process may vary from one tumor to the next, based on the heterogeneity of the microvasculature. In general, the microcirculation of tumors is characterized by low pressure (26), regurgitant flow, intermittent stasis, vessel tortuosity, spontaneous hemorrhage, and regions of low p02 (10,36, 37). It is known that these characteristics are both pathophysiologic and heterogeneous, when compared with normal tissues, and that they are susceptible to temporal fluctuations (4, 37). The purpose of this study was to explore potential

METHODS

AND

MATERIALS

Animal model Fischer-344 rats with transparent window chambers transplanted into the dorsal skin flap were used to visualize

Presented in part at the 36th Annual Meeting of the Radiation

This work was supported

Research Society, Philadelphia, April 17, 1988. Reprint requests to: Mark W. Dewhirst, D.V.M., Ph.D., PO Box 3455, Duke University Medical Center, Durham, NC 27710.

Accepted

559

for publication

by PHS grant # ROl CA 40355-03. 9 August 1989.

560

1. J. Radiation Oncology 0 Biology 0 Physics

granulating subcutaneous tissues and to provide a substrate for tumor growth. Details of the chamber design and surgical technique have been published elsewhere (24). Briefly, aseptic surgical dissection of a 1-cm diameter hole was made in opposing surfaces of the dorsal flap. The fascia was dissected away, leaving a single fascial plane with its associated artery/vein pair(s). The two halves of the chamber were placed on either side of the resultant tissue window. Glass coverslips in the center of each chamber half provided a transparent barrier to infection and dehydration. The chamber was sutured into place and rested upright on the back of the animal. The thickness of the tissue plane was 150 pm. The tumor used for these studies was the R3230 AC mammary adenocarcinoma, which is transplantable in the Fisher-344 rat ( 14). Transplants were made by placing a O.l-mm’ piece of tumor into the center of the window chamber. Within 7 days, tumor neovascularization could be observed. Following the surgical implantation, the animals were housed individually in an environmental room maintained at 35°C 50% humidity. During experimental procedures, the animals were anesthetized with pentobarbital sodium (40 mg/kg, ip) and placed in lateral recumbency on a microscope stage. The animals were covered with reflective space blanket material to aid in maintaining body temperature. Total measurement time was not allowed to exceed 90 min to minimize artifacts in peripheral flow that might arise from prolonged anesthesia. Measurement c!fvtwel intraluminal diameter and red cell velocitJ~ Measurements of vessel intraluminal diameter and red cell velocity were done using video images obtained from transillumination of the preparation at 200X. Diameter measurements were made using a video signal passed through an image shearing monitor.* The servo-null dual spot correlation technique was used to measure red cell velocity (34).‘.t In each experiment, red cell velocities and diameters were measured in 30 to 40 vessels. The vessels were randomly selected within diameter strata, to ensure that representative measurements were made at each level which was from of branching (1 1) within the preparation, 10 to 200 pm. Measurement ofwssel dimensions,fkm photomontage’s Photomontages were made by taking overlapping picture frames to cover the entire window area. The magnification factor was approximately 70X. Percent vascular volume was estimated by using a point counting technique similar to that described by Chalkley et al. (2, 3). Briefly, the percent vascular volume was calculated from the ratio of the number of grid intersections that overlaid vessels to the total number of grid intersections seen to cover the * IPM, Model 907 shearing monitor: La Jolla, CA. + IPM, Model 204 video analyzer; La Jolla, CA. * IPM, Model 102B velocity tracker; La Jolla, CA.

March IWO. Volume 18. Number 3

preparation. Whenever a vessel passed underneath an intersection point, its diameter and length were also measured. Diameters were measured using a digital micrometer.” Length was measured using a cursor interfaced with a digitizing tablet ** to a microcomputer. ‘+ Vascular length was defined as the length of a vessel segment between two contiguous branch points. The edge of the tumor preparation was defined as being equivalent to the outside of the visual margin of the peripheral hypervascular zone (6).

Tumor tissues were transplanted into the window chamber 3 days after the window chamber itself was surgically placed into the dorsal skin flap. Photomontages and measurements of red cell velocity were made on days 14. 15, and 17 after the window chamber was surgically implanted. On day 14, the tumors measured 3 to 5 mm in diameter. Tumors were randomly assigned to the control or irradiated groups. On day 14, measurements as described above were made first. followed immediately by either sham (control) or 5-Gy (experimental) radiation exposure. ~hlculation of vascular density, intercapillar~~ spucing, and pcyfiuion The diameter of each preparation was 1 cm. and the thickness was 150 pm. Thus, the total volume of tissue in each prep was calculated to be I. I8 X IO”’ pm3. using the formula for the volume of a cylinder. The total volume of the prep occupied by tumor was determined by morphometrically determining the surface area occupied by tumor. The volume of each preparation occupied by vessels was estimated by multiplying the percent vascular volume times the total tumor volume: Vessel volume

= (%I vascular

volume)

(prep volume).

Median individual vessel volumes were calculated, assuming the vessels were cylindrically shaped. using median diameters (MD) and lengths (ML). Thus, the total number of vessels per preparation could be determined by dividing the median vessel volume into the total volume occupied by vessels. Vascular density in vessels/mm’ was then calculated by dividing the total number of vessels into the total preparation volume which was occupied by tumor. vessel density

= VD (Prep. _

volume

in mm’)(%

(MD/2)*

- x - ML

prep

volume

in mm3

e Brown and Sharpe, digit-Cal II: N. Kingstown. ** GTCO Corporation. ++IBM-PC AT.

VV)

Model 5A; Rockville,

RI. MD.

Tumor

microvascular

response

Intercapillary spacing (CS) was estimated, assuming all vessels were parallel, by dividing the median vessel diameter by the fractional vascular volume (6). Tissue perfusion was estimated using a formulation described by Intaglietta and Zweifach (17). In this model, perfusion can be calculated from:

where RBCV = median red cell velocity (in cm/set) and 6000 is a conversion factor to get to mL/ 100 g/min. This model assumes (a) that a given volume of blood passes only once through a vessel as it traverses from the inflow to outflow portion of the vasculature; (b) each vessel is associated with a volume of tissue which quantitatively is described by a cylinder whose diameter is that of the intercapillary spacing and whose length is the average capillary length: (c) the perfusion of the cylinder of tissue is determined by the flow rate of blood through the capillary that traverses it; and (d) the density of the tissue is 1 gm/cm3. In a previous publication, we have described considerable intratumoral variability in vascular diameter and length (6). Thus, these calculated values only globally represent microvascular characteristics of each tumor and do not account for spatial variability within each tumor. Nevertheless, the data can be used to broadly compare vascular morphometry between individual tumors (i.e.. intertumoral comparison) as well as to make repeated observations of the same tumor over time.

Predicted vs. mea.wred.flow changes The Hagen-Poiseuille equation ( 18) was used to predict relative changes in average flow of each tumor using changes in morphometric measurements of median diameter and length relative to pre-irradiation data. The flow rate in a tissue is given by

to radiation

0 M. W. DEWHIRST el al.

Thus, relative changes in flow were predicted by examining the ratio of relative change in (median radius)4 divided by relative change in median length. These calculations were made on individual tumors for 24 hr and 72 hr vs pretreatment data. The calculations assumed that Ap and 7 did not change during the observation period. Actual changes in flow were calculated from the ratio of median vessel flows at 24 and 72 hr divided by pretreatment data. Our previous studies have shown that there is no significant relationship between diameter and velocity in this tumor (6). Therefore, median velocities should be a reasonable descriptor of overall velocity within a tumor.

Statistical considerations Descriptive statistics were used to summarize often voluminous data from an individual tissue; the Wilcoxon rank-sum test was used to test the hypotheses of no difference in summary measures between the experimental groups and Spearman rank correlation was used to measure association. The distributions of diameter, length, and red cell velocity were far from Gaussian. For example, there was no obvious single transformation which could transform simultaneously all observed distributions of vessel diameter to approximate normality. Therefore, quartiles were used to summarize the distributions. Specifically, the measures of location and dispersion were the median and the interquartile range. The randomization model was more appropriate than the population model in interpreting the significance level of the statistical tests because there might have been unintentional biases in sample selection, particularly with respect to red cell velocity. A nonparametric test for temporal trends was used to examine whether measured and calculated parameters were changing over time within control and experimental groups (22). Statistical calculations were made using a statistical package*$ on a microcomputergO and a minicomputer.***

Q=$,

RESULTS

where Q = flow rate, Ap is the pressure difference between arterial and venous ends of the circulatory bed, and FR = flow resistance. Flow resistance is controlled by viscosity (r]), length (L), and radius (R), as follows FR

=

8sL aR4 ’

Therefore, substituting the Hagen-Poiseuille into the first equation shows that _

SAS statistical BBIBM PC. H

package.

561

ApR4

equation

On the first measurement day, median diameters ranged from 41.4 to 84.9 pm, with an average of 64.5 + 19.5 pm (Table 1). Median lengths ranged from 292 to 655 pm, with an average of 505 & 149 pm. Percent vascular volume and red cell velocities averaged 2 1.O * 6.3% and 0.53 & 0.1 1 mm/set, respectively. All tumors showed an increase in percent vascular volume, going from a presham treatment average of 2 1% to 29. I % at 72 hr later (test for trend, p = 0.0007). Statistically significant changes in measured parameters were seen in individual tumors as a function of time. However, the direction in which the changes occurred was not consistent between tumors. ***

DEC VAX.

I. J.

562

Radiation Oncology 0 Biology 0 Physics Table 1. Measured

Expt. no. 117

118

119

120

121

o+ 24 72 0 24 72 0 24 72 0 24 72 0 24 72

Average 0 24 72

morphometric

March 1990. Volume 18. Number 3 and hemodynamic

data (controls) Velocity* (mm/set)

MD (p)

ML (II)

% vv

84.9$ 91.6 138.5 41.4 37.9 51.4 81.9 67.2 42.2 48.1+ 84.3 50.2 66.11 42. I 47.6

410 430 410 292* 465 327 655* 586 393 600 589 602 567+ 460 433

14.5 15.9 18.5 24.5 22.1 30.7 29.9 33.7 42.2 15.9 16.1 33.1 20. I 17.4 20.9

0.59 0.78 0.76 0.47 0.61 0.59 0.36* 0.37 0.37 0.64 0.64 0.64 0.57’ 0.60 0.37

64.5 +- 19.5 64.6 f 24.2 65.9 -t 40.7

505 * 149.0 506 t- 75.6 433 2 102.0

21.0 f 7.5 21.0 & 7.5 29.1 -+ 9.6

0.53 * 0.15 0.60 f 0.15 0.61 f 0.15

* Median velocity; obtained from measurement of 25 to 40 vessels. + Time of measurement, relative to sham irradiation. MD = Median diameter, obtained from 100 to 300 vessels; ML = Median using a point count technique, as described in the text. z Change over time was significant (p < 0.05). Kruskal-Wallis test.

Thus, tests for trends in MD, ML, and velocity were not significant. Vascular densities ranged from a minimum of 67 to a maximum of -600 vessels/mm’, at the first timepoint (Fig. 1). Density was unstable and unpredictable in these control tumors over time. In two tumors (1 19, 12 I), densities increased over time, which was a result of decreasing MD and/or ML, coupled with an increase in percent vas-

118

MLf

121 MD,,

vascular

volume,

measured

cular volume. In experiments 1 18 and I 17 densities decreased, even though percent vascular volumes were increasing. This was caused by increases in MD or ML. The test for a trend in density was not significant (p = 0.442). Tissue perfusion rates increased over the 72-hr time period, although at 24 hr, the trend was less apparent (p = 0.045; Fig. 2). In 4 of 5 tumors, predicted change in flow, based on Hagen-Poiseuille’s equation, agreed well with measured changes in flow (Fig. 3).

In this experimental group, MD ranged from 22.9 to 7 1.1, with an average of 44.7 f 17.5 pm, prior to irradia-

119 ML1

300

length: % VV = percent

ML1

120 MD1

100 Trend toward change in perfusion vs. time

30

pm.045

117 MD t

J

72

24 Time (hours)

Fig. I. Vascular density (vessels/mm3) over 72-hr time period: control tumors. Vascular densities were variable in control tumors prior to sham radiation (time 0 hr), ranging from 60 to 600 vessels/mm3. In addition, they were temporally unstable, with some tumors tending toward an increase and others toward a decrease over time. Experiment numbers are listed on the right side of the figure. The prominent (statistically significant) measured parameters which contributed to the calculated change in density are enumerated to the right of the experiment number (from Table 1). Densities were calculated from measured data as described in the text.

0

24

72

Time (hours)

Fig. 2. Tissue perfusion over 72-hr time period: control tumors. In 4 of 5 tumors studied, perfusion increased over the observation period. In the 5th tumor (Expt 121), there was little temporal change. Perfusion values were calculated as discussed in the text.

Tumor

-4 ,: 0

4

2

PREDICTED

Fig. 3. Relationship between flow in control tumors.

microvascular

predicted

to radiation

IN FLOW

and measured

change in

tion (Table 2). Median lengths ranged from 231 to 475, with an average of 327 k 88 pm at the same timepoint. Percent vascular volume and red cell velocities averaged 20.4 + 4.9% and 0.55 + 0.1 1 mm/set, respectively. As in the controls, statistically significant changes in measured parameters were seen in individual tumors as Table 2. Measured

Expt. no. 103

105

122

123

124

125

126

o+ 24 72 0 24 72 0 24 72 0 24 72 0 24 72 0 24 72 0 24 72

Average 0 24 72

MD (PO

0 M. W.

DEWHIRST

et al.

563

a function of time, although they were not consistent for MD, ML, or velocity. Six of the seven tumors showed an increase in percent vascular volume. The pretreatment average was 20.4 -t 4.9% and increased to 25.5 + 7.9% 72 hr later (test for trend was significant, p = 0.040). Vascular densities ranged from 163 to 2800 vessels/ mm3, prior to irradiation (Fig. 4). Densities increased in six tumors and decreased in one tumor (# 126). In the six cases in which density increased, it was accompanied by a reduction in MD and/or ML. Thus, the vessels, in general, were becoming smaller. In one case (# 126), MD became larger and percent vascular volume decreased (Table 2) thus leading to a reduction in density. A test for trend in density in the irradiated group was not significant (p = 0.1 12). Tissue perfusion rates increased over the 72-hr time period in 6 of 7 tumors, but decreased in one case (# 126) (Fig. 5). The degree of change was variable between individuals and a test for trend in this group was not significant (p = 0.1 12). In 4 of 6 tumors, predicted changes in flow were high, relative to measured changes (Fig. 6).

10

6 CHANGE

response

Irradiated vs. control groups In control tumors, increases in vascular density or perfusion occurred at least one time point in all five tumors studied (Fig. 7). However, the increases were not usually data (irradiated

ML (CL)

preps) Velocity* (mm/set)

%VV

62.2+ 51.1 59.7 36. I 39.9 32.6 34.4 34.2 52.1 50.6 52.2 71.1% 26.9 31.3 36.0’ 30.6 23.3 22.9* 26.2 26.2

283+ 256 365 3414 203 475+ 224 414 329* 188 321 233+ 188 173 394’ 350 254 231 196 211

14.0 32.7 32.4 18.3 19.1 18.5 16.4 19.8 25.5 32.9 34.2 15.8 17.5 21.1 23.8 26.4 34.9 26.6 18.9 16.8

0.50’ 0.34 0.54 0.71* 0.70 0.87 0.58 0.57 0.61 0.56 0.59 0.61 0.61+ 0.19 0.38 0.34 0.26 0.34 0.57 0.60 0.55

44.7 f 17.5 36.6 + 11.4 38.1 f 13.5

327 ? 88.0 234 * 63.0 277 * 90.9

20.4 + 4.9 24.1 f 7.6 25.5 f 7.9

0.55 ?I 0.11 0.46 * 0.20 0.56 f 0.17

* Median velocity; obtained from measurement of 25 to 40 vessels. + Time of measurement, relative to irradiation. MD = Median diameter, obtained from 100 to 300 vessels; ML = Median using a point count technique, as described in the text. * Change over time was significant (p < 0.05) Kruskal-Wallis test.

length; % VV = percent

vascular

volume,

measured

J.

1.

564

Radiation

Trend toward tdenslly

loooo

1990, Volume 18. Number 3

Oncology 0 Biology 0 Physics

p=.,W

? 125 MDl. ML1

103 MDL, MLlt

100

I

’

24

0

Time (hours)

72

Fig. 4. Vascular density (vessels/mm3) over 72-hr time period: irradiated tumors. Vascular densities were also variable in ir-

/ f

radiated tumors prior to irradiation, varying by over two orders of magnitude. In most ofthe tumors. vascular density increased. although in Experiment 126 a reduction in density was observed. Experiment numbers are listed on the right side of the figure.

0.0

0.5

1.0

2.0

1.5

PREDICTED

CHANGE

Fig. 6. Relationship between predicted tumors. flow in irradiated

conjoint (i.e.. density and perfusion changed in same direction at a fixed time point). In irradiated tumors, however, density and perfusion usually changed conjointly, although there was considerable heterogeneity in the degree of the effect. In the irradiated group, the tumor that had the largest pretreatment median diameter (# 124; 7 1.1 pm) had the largest increase in density at 24-hr post-irradiation. whereas the tumor with the smallest pretreatment diameter (#126; 22.9 pm) had a slight decrease in density (Fig. 8). A triexponential fit to the data comparing pretreatment diameter to relative density change was significant (R* = 1.000). Thus, there appeared to be a direct relationship between density change and pretreatment morphology, as described by median diameter. There was no discernable relationship between pretreatment diameter and density change in the control group, however.

2.5

3.0

IN FLOW

and measured

change

in

compatible with reoxygenation, while in others they were not. The major question which needs to be addressed, therefore, is what is the reason for the heterogeneity? The results of this work point to two potential sources. First, the vascular morphology prior to radiation apparently has some influence. This was demonstrated in the positive correlation between pretreatment diameter and density change. The second clue comes from the deviation between predicted and measured flow changes. The HagenPoiseuille equation predicts that flow is dependent on morphology (which was measured) as well as the pressure drop and viscosity (neither was measured in this study). Irradloted

Group

Control

Group

DISCUSSION

Hrtercgeneit~~ in vascular events ,Ji,llm~ing irradiation The results of this study reveal a wide range of effects from radiation exposure. In some tumors, the effects were

l . ..

* 2.0 9 E

1.5 -

(241 (72)

II9

(24) (72

123

(241 (72)

120

124) 1721

124

125

2.5 I .:

122

,.’ *.** ,..* /

*...*.... .-..

A.._

*.._. .._

*%.A

I

;;I

(24) (721

Scale 123 .

Fig. 5. Tissue perfusion over 72-hr time period: irradiated tumors. In irradiated tumors, changes in perfusion were variable from one tumor to the next, but in all tumors except one (# 126) perfusion increased, at least to a small degree over the observation period.

I

.....

t-4 : 100% = density : perfwon

Fig. 7. Relative changes in density and perfusion for control and irradiated groups. The relative changes in density and perfusion are depicted as bars, originating from a zero change vertical ordinate. When the bars point to the left of the ordinate, the direction change is negative. When the bars point to the right, the direction of change is positive. The data are displayed at 24 and 72 hr, relative to time 0 hr, for each experiment. Changes in perfusion and density for each irradiated tumor were often conjoint (i.e., same direction), although the magnitude of the change varied considerably. In control tumors, the changes in perfusion and density were not usually conjoint.

Tumor

microvascular

response

E 5

8--

7--

A-A

Controls

-24

o---e

Irradiated-24hrs

hrs

6--

+-

5--

z

4--

f .P ‘0 u

3-2__

e

o--

l--l-

0

I IO

1 20

, 30

Diameter

40

0 M. W. DEWHIRST el al.

565

lary permeability has been shown to increase by a number of investigators ( 12, 2 1).

J

zk

to radiation

50

60

70

80

at Time 0 Hours

Fig. 8. Relative change in density at 24 hr as a function of median pretreatment diameter. These results are consistent with the hypothesis that one overall effect of radiation on tumor microvasculature is to reduce vessel diameter. In tumors with the largest sinusoidal vessels (-70 pm diameter), radiation resulted in an 8 to 9-fold increase in density, concomitant with a reduction in median diameter (Table 2).

Thus, when a discrepancy existed between predicted and measured changes in flow, it can be deduced that changes in pressure or viscosity influenced the results. In this study, it was not possible to definitely determine whether radiation was entirely responsible for the deviations between predicted and measured data since deviations were observed in both the irradiated and control groups. However, the one discrepancy in the control group may have been due to sampling errors. In that particular case, the morphologic data predicted nearly a IO-fold increase in flow. In order to see that large a change in flow, the velocity would have to increase by nearly the same amount. Baseline red cell velocities tended to be in the range of 0.4 to 0.6 mm/set (Table 1). The system used for these studies could only measure velocities up to 1.5 mm/set; thus, it was incapable of detecting this magnitude of change. On the other hand, the magnitude of changes predicted in all tumors of the irradiated group were within a factor of two, and thus were within the measurement limits of the system. Based on these considerations, the data imply that radiation may cause changes in pressure and/or viscosity that affect vascular responses. Changes in blood pressure could occur as a result of radiation effects in the normal tissue bed surrounding the tumor. Prior work in this laboratory (5) and others (7, 8) has shown that irradiation creates a loss in capillary density in normal tissues. The effects are most pronounced in the smallest diameter vessels, implying that shunts must open to bypass the defunctionalized vessels. This hypothesis is supported by the observation that red cell velocity tends to increase following radiation exposure (5). In this study, both the tumor and its surrounding normal tissue bed were irradiated. Thus, the creation of shunts in the normal tissue could indirectly effect the pressure drop within the tumor. Changes in viscosity could occur as a result of inflammatory events following irradiation. For example, capil-

Another explanation for the discrepancy between the predicted and measured flow changes may be related to endothelial swelling and hypertrophy that has been documented to occur as soon as 24 hr after irradiation (12, 21). Recent studies by Rosen et al. (31) have shown that cell loss occurs even in nonproliferating, confluent plateau-phase cultures of endothelial cells. Furthermore, the remaining cells attempt to compensate for the lack of confluence by synthesizing protein and increasing cell volume. In our studies on tumors, diameters as measured from photomontages represented outside diameters. Thus, it is not possible to determine whether intraluminal diameters were changing. If they become smaller as a result of endothelial swelling, then the morphometric measurements could have easily overpredicted changes in flow. Mechanisms

qj’reoxygenation

The process of reoxygenation, as described by Kallman ( 19) refers to the re-acquisition of radiosensitivity (by reacquisition of oxygenation) by formerly hypoxic survivors of a given radiation exposure. This process has been studied using a variety of indirect in vivo techniques, including paired survival curves, clamped tumor control, and clamped growth delay (20). Using these types of assay systems, most tumors show evidence for reoxygenation within 6 to 24 hr following exposures of 10 to 15 Gy, although, in some tumor lines, the process appears to be incomplete, even at 13 days after a conditioning exposure of 15 Gy. When radiation doses are fractionated, the hypoxic fraction at the end of the course is the same as in untreated tumors, indicating that fractionated therapy may be more efficient at inducing reoxygenation than single doses (20). A number of mechanisms have been proposed for the process of reoxygenation, including post-radiation reduced 02 metabolism, improved circulation, tumor cell loss, and cellular migration (19). The rapidity with which the process occurred in this study tends to make the latter two mechanisms less plausible, at least shortly following a single dose of irradiation. Reinhold et al. (28) studied the process indirectly by fluoroscopically examining the conversion of NAD to NADH induced by hypoxia in a dorsal flap window chamber containing a tumor. An increase in the fluorescence response at 24 hr following an exposure of 20 Gy implied that more oxygen was available than prior to radiation. Since the marker used is a metabolite of cellular respiration, the results tended not to support the hypothesis that reduced cellular respiration contributed to reoxygenation from a single radiation dose. More recent studies by Tozer et a/., examining energy metabolism in irradiated tumors, demonstrated that lactate concentrations dropped 20 hr following single exposures of 20 Gy and that the drop was accompanied by increases in tumor perfusion and AMP (35). When fractionated doses were given, blood perfusion rates remained constant,

566

I. J. Radiation

Oncology

0 Biology l Physics

but phosphorus metabolite levels increased while lactate levels decreased. The conclusions of this study were comparable to prior studies suggesting that, following single doses of radiation, reoxygenation may be related to vascular events. During a course of fractionated therapy, parenchymal cell loss could also play a role (4, 33). The results of this study are in agreement with a number of previously published reports. One of the earliest reports was by Rubin and Casarett, who, using microangiographic techniques, described a “supervascularized” state in murine tumors following single and fractionated doses of radiation (32). Hilmas and Gillette described a similar phenomenon in irradiated C3H mammary tumors, using histomorphometry (15, 16). Reinhold et ul. studied vascular function in a dorsal flap window chamber, containing the CH3 HBA tumor, following fractionated radiotherapy by measuring integrated vascular length at defined intervals following injection of a fluorescent dye (27). The rate of change of this end point after dye injection would be dependent on capillary density and/or blood velocity. The results indicated approximately a 30% increase in integrated vascular length after 5 to 6 days of daily fractions of 5.7 Gy each. Thus, in that model it appeared that vascular effects persisted after several days of fractionated radiotherapy. Relationship between vuscuiur morphology and prognosis,fiom rudiotherapl Heterogeneity in tumor microvasculature has been the subject of numerous investigations. Whereas it is possible to generally associate certain types of vascular patterns with specific histologic types (37), it is known that there is considerable heterogeneity within a group of tumors of the same histologic type, as well as within a single tumor (10, 1 1, 37). Intertumoral variability has been thought to be at least partially responsible for altering the slope of radiation dose-effect curves (30). Clinically, there is evidence that variability in human tumor oxygenation exists, both within and between patients, even when the histologic type is the same (13, 23. 25). Furthermore, the variability may have prognostic value for treatment with radiotherapy. Revesz and Siracka examined the relationship between percent vascular volume in pretreatment biopsy specimens and prognosis in two groups of age- and stagematched patients with carcinoma of the cervix that were treated with radiotherapy (29). The patients who survived 5 years or longer tended to have a larger percent vascular volume than patients who survived less than 5 years, indicating that tumor vascularity might be related to radiocurability. More recently, Awwad et al. evaluated mean tumor intercapillary distance as a prognostic indicator in patients with Stage IIB and III carcinoma of the cervix (1). The overall average intercapillary distance was found to be 304 t- 30 pm. The mean intercapillary distance was larger for patients who had local failure following radiotherapy, as compared with those who had local control. This prog-

March

1990. Volume

18. Number

3

nostic indicator was independent of clinical stage and degree of differentiation, thus indicating that it might be a useful tool for identifying patients in which hypoxia may interfere with radiocurability. In the present study, intercapillary spacing was estimated by dividing the median diameter by the fraction of tumor occupied by vessels (6). In the irradiated group of tumors prior to radiation exposure, experiment 124 had the largest calculated spacing of 450 pm, while experiment 126 had a spacing of 86 pm (Table 2). Thus, these data are in rough agreement with capillary spacing measurements that have been made in human patients. Recently. Gatenby et al. (13) studied the relationship between tumor oxygenation, as measured by needle electrodes, and prognosis in human patients with squamous cell carcinoma metastases. They were able to demonstrate that patients with greater than 26% of the tumor volume < 8 mm Hg had a much worse prognosis than patients who had less than 26% of their tumor volume hypoxic. These results strongly suggest that tumor blood flow and oxygenation prior to therapy have a strong influence on treatment outcome. In this study, intertumoral variabilities in vascular density and perfusion rates were quite large, being nearly three orders of magnitude for density and a factor of six for perfusion rates. Given the degree of variability seen in these tumors, it would be likely that a large range in hypoxic fraction exists as well. Compurison of irrudiuted vs. control tumors prior to treatment Even though the animals were randomly assigned to the control or experimental groups, it is clear that the groups were not entirely comparable at the first time point (Tables 1 and 2). For example, MD averaged 20 pm smaller in the irradiated than the control group (two-tailed t-test for comparison of means p = 0.1). Similarly, ML averaged 327 pm and 505 Frn in the irradiated and control groups, respectively (two-tailed t-test for comparison of means; p < 0.05). Thus, vascular densities and perfusions were higher in the irradiated than the control group, although there was some overlap (Figs. 1 to 4). Since the groups were somewhat different at the beginning of the experiment, interpretation of the results, regarding evidence for treatment effects, needs to be cautious. This was further complicated by the temporal instability in measured vascular parameters that was encountered in the control group. In irradiated tumors, the conjoint changes in density and perfusion confirm that irradiation caused a measurable effect, relative to the control group. In the controls, changes were observed but they were not usually conjoint. Calculation @“petjiuion ,fiom microvascular measurements The absolute values of perfusion, which were calculated from the morphometric and dynamic flow data, are sub-

Tumor microvascular response to radiation 0 M. W. DEWHIRST et ul.

ject to a number of sources of error which make them much higher, and therefore not directly comparable, to other techniques for measuring perfusion. Some of the assumptions were listed in Methods and Materials. For example, in tumors, the assumption that the blood enters one side of the tissue and passes through only once is most certainly erroneous, especially considering the disorder of the vascular bed, which has regurgitant flow with little or no arteriolar input. The second assumption, which may be in error, is that each vessel contributes to the overall perfusion in a physiologically significant way (i.e., the perfusion of a cylinder of tissue is determined by the flow rate of blood through the capillary that traverses it). In tumors, it is likely that many vessels with low flow and/ or low pOz do not contribute significantly to the delivery

567

of nutrients to the surrounding tissue. Finally, the estimates of capillary density in this study were based on morphometric measurements made from photomontages, which made it impossible to distinguish flowing and nonflowing vessels. Perfusion, calculated from morphometric and flow measurements, has been compared with more traditional techniques such as microspheres (17). In general, the former tends to give higher values than the latter, except in the case of very highly perfused tissues, such as heart muscle. For these reasons, the perfusion values reported in this study should be used only in a comparative sense, as their absolute accuracy is questionable. Since each animal served as its own control over time, temporal comparisons are reasonable.

REFERENCES 1. Awwad. H. K.; el Naggar, M.; Mocktar, N.; Barsoum. M. Intercapillary distance measurement as an indicator of hypoxia in carcinoma of the cervix- uteri. ht. J. Radiat. Oncol. Biol. Phys. 12:1329-1333; 1986. 2. Chalkley, H. W. Method for the quantitative morphologic analysis of tissues. J. Natl. Cancer Inst. 4:47-53; 1943. 3. Chalkley, H. W.; Cornfield. J.; Park, H. A method for estimating volume-surface ratios. Science 110:295-297; 1949. 4. Clement, J. J.; Tanaka, N.; Song, C. W. Tumor reoxygenation and postirradiation vascular changes. Radiology 127: 799-803; 1978. 5. Dewhirst. M. W.; Gustafson, C.; Gross, J. F.; Tso. C. Y. Temporal effects of 5.0 Gy radiation in healing subcutaneous microvasculature of a dorsal flap window chamber. Radiat. Res. I12:581-591; 1987. 6. Dewhirst, M. W.; Tso, C. Y.; Oliver, R.; Gustafson, C. S.; Secomb. T. W.: Gross, J. F. Morphologic and hemodynamic comparison of tumor and healing normal tissue microvasculature. Int. J. Radiat. Oncol. Biol. Phys. l7:91-99; 1989. 7. Dimitrievich, G. S.; Fischer-Dzoga, K.; Griem, M. L. Radiosensitivity of vascular tissue. Radiat. Res. 99:5l l-535; 1984. 8. Dimitrievich, S. L.; Hausladen, F. T.; Kuchnir. F. T.; Griem, M. L. Radiation damage and subendothelial repair to rabbit ear chamber microvasculature: an in viva and histologic study. Radiat. Res. 691276-292; 1987. 9. Dorie, M. J.; Kallman, R. F. Reoxygenation of the RIF-1 tumor after fractionated radiotherapy. Int. J. Radiat. Oncol. Biol. Phys. 12: 1853- 1859: 1986. 10. Eddy, H. A.; Casarett, G. W. Development of the vascular system in the hamster malignant neurilemmoma. Microvascular Res. 6:63-82; 1973. Il. Endrich, B.; Reinhold, H. S.; Gross, J. F.; Intaglietta, M. Quantitative studies of microcirculatory function in malignant tissue: influence of temperature on microvascular hemodynamics during the early growth of BA 1112 rat sarcoma. Int. J. Radiat. Oncol. Biol. Phys. 5:2021-2030; 1979. 12. Eriksson, B.; Johnson, L.; Lundqvist, P-G. Ultrastructural aspects of capillary function in irradiated bowel: an experimental study in the cat. Stand. J. Gastroenterol. 18:473480; 1983. 13. Gatenby, R. A.; Kessler, H. B.; Rosenblum, J. S.; Coia, L. R.; Moldofsky, P. J.; Hartz, W. H.; Broder, G. J. Oxygen distribution in squamous cell carcinoma metastases and its relationship to outcome of radiation therapy. Int. J. Radiat. Oncol. Biol. Phys. 14:831-838; 1988.

14. Hilf, R.; Michel, 1.; Bell, C.: Freeman, J. J.; Borman, A. Biochemical and morphologic properties of a new lactating mammary tumor line in the rat. Cancer Res. 25:286-299; 1965. 15. Hilmas, D. E.; Gillette, D. E. Microvasculature of C3H/Bi mouse mammary tumors after x-irradiation. Radiat. Res. 61:128-143; 1975. 16. Hilmas, D. E.; Gillette, E. L. Tumor microvasculature following fractionated x irradiation. Radiology I 16: I65- 169; 1975. 17. Intaglietta, M.; Zweifach. B. W. Microcirculatory basis of fluid exchange. Adv. Biol. Med. Phys. 15: 112- 159; 1974. 18. Jain, R. K. Determinants of tumor blood flow: a review. Cancer Res. 48:2641-2658; 1988. 19. Kallman, R. F. The phenomenon of reoxygenation and its implications for fractionated radiotherapy. Radiology 105: 135-142; 1972. 20. Kallman, R. F.; Dorie, M. J. Tumor oxygenation and reoxygenation during radiation therapy: their importance in predicting tumor response. Int. J. Radiat. Oncol. Biol. Phys. 12:681-685; 1986. E. Effects of radiation on microcirculation. 2 I. Kivy-Rosenberg, In: Kaley, G., Altura, B. M.. eds. Microcirculation, Vol. III. Baltimore, MD: University Park Press; 1980:5-20. 22. Lehman, E. L. Nonparametrics: statistical methods based on ranks. New York: McGraw Hill; 1975:290-297. 23. Mueller-Klieser, W.; Vaupel, P.; Manz, R.; Schmidseder, R. Intracapillary oxyhemoglobin saturation of malignant tumors in humans. Int. J. Radiat. Oncol. Biol. Phys. 7: 13971404; 1981. 24. Papenfuss, D.; Gross, J. F.; Intaglietta, M.; Treese, F. A. A transparent access chamber for the rat dorsal skin fold. Microvas. Res. 18:31 l-318; 1979. 25. Pappova, N.; Siracka, E.; Vacek, A.; Durkovsky, J. Oxygen tension and prediction of the radiation response. Polarographic study in human breast cancer. Neoplasma 29:669674; 1982. 26. Peters, W.; Teixeira, M.; Intaglietta, N.; Gross, J. F. Microcirculatory studies in rat mammary carcinoma 1. Transparent chamber method, development of microvasculature and pressures in tumor vessels. J. Natl. Cancer Inst. 65: 631-642; 1980. 27. Reinhold, H. S. The response of tumor blood supply to fractionated irradiation. Conference on Time and Dose Relationships in Radiation Biology as Applied to Radiotherapy. Carmel, CA: Sept. 15- 18, 1969:2 1O-2 14.

568

1. J. Radiation Oncology 0 Biology 0 Physics

28. Reinhold, H. S.; Blachiewicz, B.: Berg-Blok, A. Reoxygenation of turnout-s in ‘sandwich” chambers. Europ. J. Cancer 15:481-489; 1979. 29. Revesz, L.; Siracka, E. Tumor vascularization, hypoxia, staging of tumors and radiocurability. Strahlentherapie 160: 658-660; 1984. 30. Rockwell, S.: Moulder, J. E.; Martin, D. F. Tumor-to-tumor variability in the hypoxic fractions of experimental rodent tumors. Radiother. Oncol. 2:57-64: 1984. 3 I. Rosen, E. M.: Vinter, D. W.: Goldberg, I. D. Hypertrophy of cultured bovine aortic endothelium following irradiation. Radiat. Res. 117:395-408: 1989. 32. Rubin. P.; Casarett, G. Microcirculation of tumors. Part II: the supervascularized state of irradiated regressing tumors. Clin. Radio]. 17:346-355; 1966. 33. Song, C. W.: Payne. J. T.: Levitt, S. H. Vascularity and

March 1990, Volume

34.

35.

36.

37.

18. Number

3

blood flow in x-irradiated walker carcinoma 256 of rats. Radiology 104:693-697; 1972. Tompkins. W. R.; Monti, R.; Intaglietta, M. Velocity measurements of self-tracking correlator. Rev. Sci. Instru. 45: 647-649: 1974. Tozer, G.: Suit, H. D.: Barlai-Kovach, M.; Brunengraber, H.; Biaglow, J. Energy metabolism and blood perfusion in a mouse mammary adenocarcinoma during growth and following x irradiation. Radiat. Res. 109:275-293; 1987. Vaupel, P.: Frinak. S.; O’Hara, M. Direct measurement of reoxygenation in malignant mammary tumors after a single large dose of irradiation. Adv. Expt. Med. Biol. 180:773782; 1984. Warren, B. A. The vascular morphology of tumors. In: Peterson, H. I., ed. Tumor blood circulation. Boca Raton: CRC Press; 1978: l-47.