Tumor Oxygenation in Hormone-Dependent Tumors During Vascular Endothelial Growth Factor Receptor-2 Blockade, Hormone Ablation, and Chemotherapy¹

Solid tumors are typically characterized by hypoxia(1, 2, 3). Despite continuous angiogenesis, local disparities between the rate of oxygen supplied by the blood and the oxygen consumption rate(Q_O2)³of the tissue leave large numbers of cells in a state of oxygen deprivation. Some of these hypoxic tumor cells remain viable and contribute to the resistance of tumors to radiotherapy and to some anti-neoplastic drugs. As a result, therapies that affect the tumor environment have become a major focus in cancer research (3, 4). The hypoxic tumor microenvironment is a stimulus for angiogenesis, up-regulating expression of angiogenic factors such as vascular endothelial growth factor (VEGF) by tumor (5) and stromal (6) cells. VEGF is a potent angiogenic factor that promotes proliferation of tumor vascular endothelium and increases vascular permeability (7, 8). Thus, therapies that neutralize or interfere with VEGF signaling should cause vessel regression and inhibit growth (e.g., Refs.8, 9, 10, 11, 12). One would also expect these therapies to increase hypoxia in tumors and limit the effectiveness of radiation and certain drug therapies. Surprisingly, the combination of angiogenesis inhibitors and radiotherapy or chemotherapy has led to additive or synergistic tumor response⁴(13, 14, 15). Similarly, hormone ablation in hormone-dependent tumors, which leads to impaired vascular function(9), has shown synergistic effects when used in combination with radiation therapy (16).

Understanding the dynamics of tumor oxygenation during tumor regression and relapse is, thus, crucial for optimal scheduling of combined treatment (1, 2). Separate measurements of the partial pressure of oxygen (pO₂) and Q_O2 have been made in tumors and, in limited cases, during response to therapy⁴ (4, 17). However,noninvasive measurements of pO₂ profiles (and calculation of Q_O2) during treatment have not been reported. Here we quantify for the first time temporal changes in tumor pO₂ and Q_O2 in response to antiangiogenic,hormone ablation, or chemotherapy (doxorubicin/cyclophosphamide). Androgen-dependent Shionogi tumors were grown in transparent dorsal skin-fold chambers in severe combined immunodeficient mice. The transparent chamber model allowed us to visualize tumor growth,regression, and relapse-over extended periods (18) and to non-invasively measure tissue pO₂ using the recently developed phosphorescence quenching microscopy (PQM) technique(2).

Male severe combined immunodeficient mice from our gnotobiotic colony were implanted with transparent dorsal skin-fold chambers(18). Androgen-dependent male mouse mammary carcinoma(Shionogi) was implanted 3 days after chamber implantation as described previously (9). Intravital microscopy was performed every third day to document tumor growth and vascularization(18). Because tumor size is an important determinant of therapeutic outcome, treatment was initiated 13 days following tumor implantation, after the tumors had become established and had developed a functional vascular network. Mice in the antiangiogenic group received either anti-VEGFR-2 monoclonal antibody (mAb; DC101, 45 mg/kg,i.p.; Ref. 19) or an equal amount of non-specific rat IgG(Jackson Immunochemicals, West Grove, PA) every 3 days. Mice in the chemotherapy group received either doxorubicin (6.5 mg/kg, i.p.) and cyclophosphamide (100 mg/kg, i.p.) or an equal amount of physiological saline every 7 days. Mice in the hormone ablation group underwent either an orchiectomy (9) or a sham operation. All procedures were carried out following approval of the Institutional Animal Care and Use Committee.

The growth data were described using the rectilinear form of the Gompertz equation (20): ln[ln A_max − ln A(t)] = lnβ/α − αt, which relates tumor cross-sectional area(A) to time (t). The parameters (α, β) were determined by comparing this model to the experimental growth data using a linear regression.

The vascular volume per unit area was calculated based on the vessel diameter (D) and length (L):

where M is the total number of vessels in the region with the area A(11). The vessel diameter and length were measured in five distinct regions (each ≈0.5 mm in length) using image analysis of the tumor vasculature.

High resolution PQM was used to measure pO₂profiles (2). Tissue pO₂ was measured between adjacent vessels in 20-μm increments. Briefly,albumin-bound palladium meso-tetra-(4-carboxyphenyl)porphyrin (Medical Systems Corp., Greenvale, NY) was injected i.v. via the tail vein (60 mg/kg). Three to seven fields of view were selected randomly throughout the tumor. After 30 min, excitation was generated by flashes (n ≤ 15) from a 540-nm flashlamp(EG&G, Salem, MA). Phosphorescence signals (≥630 nm; Oriel, Stratford,CT) were detected with a photomultiplier tube (9203B,Products for Research, Danvers, MA), averaged on a digital oscilloscope(TDS-320, Tektronix, Beaverton, OR), and transferred to a computer. Lifetime constants (τ) were obtained from this data and pO₂ values were calculated using the Stern-Volmer equation, pO₂ = 1/k(1/τ − 1/τ₀)(21). In this expression, k andτ ₀ are the quenching constant and the phospho-rescence lifetime in the absence of oxygen, respectively. To establish pre-treatment values, oxygen levels were measured on day 12,1 day before initiating therapy. After initiating therapy, oxygen levels were measured 14, 20, 26, and 32 days after tumor implantation. RBC velocity was also measured in these vessels using the temporal correlation velocimetry described elsewhere (19).

To estimate changes in Q_O2 with time, a one-dimensional linear diffusion-reaction model was used:

pO₂ data were fit with this second order polynomial. We assumed a value of D = 2 × 10⁻⁵ cm²/s (22), S = 2.05 × 10⁻⁵ml_O2/cm³ · mm Hg (23), and calculated Q_O2 (ml_O2/cm³· min). Q_O2 were calculated only for those vessels that were greater than 200 μm from the nearest neighboring vessel to satisfy the assumptions in the mathematical model.

Results are presented as mean ± SE. ANOVA tests were performed to compare the equality of controls. F-tests were used to test the equality of variances. Two sample t-tests for independent samples of equal variances were performed to compare sample means. Statistical significance was based on P values smaller than 5%. Prior to the first day of treatment (day 13), there were no statistically significant differences between the four groups(control, anti-VEGFR-2 mAb, hormone ablation, and doxorubicin/cyclophosphamide) included in the study.

Tumor size and transformed tumor size analyzed using the rectilinear form of the Gompertz model are shown in Fig. 1 and B, respectively. Tumors in the control groups exhibited similar growth curves, independent of the treatment modality (saline, non-specific IgG, or sham operation). Hormone ablation and treatment with doxorubicin/cyclophosphamide resulted in significant growth delays (P < 0.05). The growth of the anti-VEGFR-2 mAb-treated group was not significantly different from the control group. The growth data were used to ensure that comparisons of pO₂ and Q_O2 values were size-matched because tumor size is known to be an important determinant of oxygenation (24). Fig. 2 shows a representative set of pO₂ data and the corresponding fit of the data to the diffusion-reaction equation used to calculate Q_O2.

Established solid tumors with a functional vascular network were most oxygenated (mean ± SE = 22 ± 3 mm Hg) 14 days after tumor implantation (Fig. 3,A). When tumors became larger, tumor oxygenation decreased. On day 26 when control tumors reached the size of the dorsal chamber,tumor oxygenation decreased further to values of about 11 ± 2 mm Hg. In contrast to tumor pO₂, oxygen consumption rate remained constant throughout the entire experiment in control tumors (Fig. 3 E), indicating that for this tumor,oxygenation is size-dependent but Q_O2 is not.

The first effect of anti-VEGFR-2 mAb on tumor pO₂occurred within 24 h of initiating treatment. Compared with size-matched controls (22 ± 2 mm Hg), a significant(P < 0.05) drop in pO₂occurred in the anti-VEGFR-2 mAb (15 ± 1 mm Hg) group. Vessel volume per unit area remained constant during this time (Fig. 4). One week after initiation of therapy, pO₂values decreased further to 10.2 ± 1.9 mm Hg(P < 0.05, 41% decrease compared with size-matched controls; Fig. 3,B). The decrease in pO₂ was accompanied by a significant(P < 0.05) decrease in vessel volume per unit area (Fig. 4), though Q_O2 remained unaffected (Fig. 3,F). Vessel volume per unit area decreased over days 20–26 and reached a minimum after day 26, 13 days after initiation of anti-VEGFR-2 mAb treatment. At this point, vessels appeared less tortuous (Fig. 5,A). Despite the decrease in vessel volume per unit area,pO₂ levels did not further decrease below 10 mm Hg. Approximately 3 weeks after initiating anti-VEGFR-2 mAb treatment,increased vessel volume per unit area was observed (Figs. 4,5,A), accompanied by increased pO₂(137% compared with size-matched controls) and Q_O2 (Fig. 3, B and F).

Changes in pO₂ and Q_O2 following hormone ablation mirrored two distinct stages of tumor vascularization:regression and regrowth. Analogous to anti-VEGFR-2 mAb therapy,pO₂ values dropped significantly(P < 0.05) compared with size-matched controls within 24 h of hormone ablation (from 22 ± 2 to 15 ± 2 mm Hg). Tumor and vessel regression(Fig. 5,B) occurred 1 week after hormone ablation, during which there was an increase in tumor pO₂ to 23 ± 2 mm Hg (P < 0.05, 30%compared with size-matched controls; Fig. 3,D), likely the result of a decrease in oxygen demand. The maximum tumor regression(Fig. 1,A) and minimum Q_O2 (Fig. 3,H) with constant pO₂(22 ± 2 mm Hg) coincided on day 26. Tumor regression was followed by tumor relapse after day 26 accompanied by a “second wave” of angiogenesis (Fig. 5,B) and an increase in Q_O2 (Fig. 3 H). Despite the increase in Q_O2,pO₂ levels remained higher(P < 0.05, 51% compared with size-matched controls) in the treatment group.

Doxorubicin/cyclophosphamide induced tumor growth arrest without regression and did not significantly affect tumor oxygenation or Q_O2. Tumor oxygenation remained constant at 22 ± 2 mm Hg during the first week of treatment (day 14 to day 20). Two weeks after treatment, tumors exhibited a slight decrease to 16 ± 2 mm Hg (P < 0.05, 13% less compared with size-matched controls; Fig. 3,C). Q_O2 remained unchanged throughout the entire experiment (Fig. 3 G).

The importance of low oxygen in cancer treatment response has been discussed since the description of tumor hypoxia by Thomlinson and Gray(25). To improve existing therapies and to develop new strategies, it is essential to understand how various tumor therapies influence the local microenvironment. Because tumor pO₂ fluctuates, reflecting the imbalance between oxygen supply and Q_O2, a continuous, controlled, non-invasive in vivo analysis is necessary to gain insight into the dynamics of tumor oxygenation(2).

Oxygen tension in all control groups decreased with increasing tumor size despite increases in vessel density. Q_O2, on the other hand, remained constant, indicating unchanged metabolic demands. This implicates hindered oxygen delivery as the cause of tumor hypoxia. This is supported by the observation that pO₂ at the vessel wall decreases with increasing tumor size. The decrease in delivery could be attributable to increased consumption by endothelial cells (26, 27), impaired oxygen carrying capacity, or increased oxygen extraction from blood as it flows over longer distances through the growing tumor.

The decrease in pO₂ observed 1 day after initiating treatment in the hormone ablation and anti-VEGFR-2 mAb groups is likely the result of the ability of these treatments to alter vascular function by inducing endothelial cell apoptosis(9) or by blocking VEGF signaling. In both treatments, the subsequent period of low pO₂ is followed by increases in pO₂ and Q_O2,coincident with a second wave of angiogenesis. In the anti-VEGFR-2 mAb-treated group, the high pO₂ levels observed with increases in vasculature and despite increases in Q_O2 may be the result of increased tissue perfusion. Blood flow rate measurements indicate that RBC velocity during vessel regression and relapse (day 14: V_RBC = 74.9 ± 2.1 μm/s; day 20: V_RBC = 80.9 ± 1.8 μm/s; day 26: V_RBC = 79.7 ± 3.3 μm/s; day 30: V_RBC = 72.9 ± 3.1 μm/s; and day 35: V_RBC = 75.4 ± 2.1 μm/s) does not change appreciably (P > 0.23, one-way ANOVA). However, vessel density increases significantly during the later stages of therapy (after day 26),suggesting improved tissue perfusion. In addition, similar to observations made in studies in which VEGF itself was neutralized(11) or down-regulated (9), vessels that develop during the second wave of angiogenesis appear less tortuous(Fig. 5). The increase in Q_O2 is presumably due to an increase in the number of oxygen-consuming cells and mirrors the metabolic demands of the tumor. Thus, it appears that Q_O2,not pO₂, mirrors the dynamics of tumor growth and regression in agreement with the findings of Gullino et al.in tissue-isolated tumors (28).

To investigate the possibility that the second wave of angiogenesis observed during anti-VEGFR-2 mAb therapy may be the result of hypoxia-induced up-regulation of VEGF (9), anti-VEGFR-2 mAb was administered in excess to ensure complete competitive inhibition. The second wave of angiogenesis still occurred under these conditions, suggesting that it may be mediated by alternate signaling pathways or by other angiogenic factors. The mechanisms underlying the second wave of angiogenesis require further investigation.

During doxorubicin/cyclophosphamide therapy there was no significant tumor growth or change in tumor vascularization (data not shown). The constant pO₂ and Q_O2 levels observed following therapy are compatible with growth arrest during which there is little increase in the number of tumor cells. These findings are consistent with other published observations(17). The slight decrease in tumor oxygenation on day 26 is consistent with the slight increase in tumor size. Compared with size-matched controls, pO₂ did not change. The effectiveness of radiation is enhanced by increased oxygenation and/or reduced tumor volume (29). Because neither a change in Q_O2 nor an increase in pO₂ (compared with size-matched controls) was observed during chemotherapy, the benefits of combining radiation with chemotherapy are likely not the result of enhanced oxygenation but rather reduced tumor volume (30). Reduced tumor volume is likely the effect of chemotherapy on both tumor cells and endothelial cells (31, 32).

Tumor size and oxygenation are predictors of the success of radiation therapy (1) and chemotherapy (33). The elevated pO₂ resulting from the second wave of angiogenesis in anti-VEGFR-2 mAb therapy and hormone ablation should enhance the effectiveness of oxygen-dependent treatments such as radiation therapy. This finding may help to explain the beneficial effects of combining antiangiogenic and radiation therapy⁴ (4, 14). This finding also illustrates the importance of timing in the administration of combined treatments. Following hormone ablation therapy, the second wave of angiogenesis and increase in pO₂ are preceded by tumor regression that is most pronounced 12–14 days after treatment. The effectiveness of radiation therapy is therefore likely to be enhanced just before relapse when tumors exhibit increased pO₂ and decreased size. Indeed, Zietman et al. (16) observed radiation therapy to be most effective 12–14 days after hormone ablation. Collectively, these findings point out that understanding the temporal dynamics of tumor oxygenation is necessary in guiding the schedule of combined therapies. The development of PQM allows the temporal dynamics of tumor oxygen tension to be measured non-invasively and chronically. If similar data are used in the clinic to develop optimal scheduling of combined treatment modalities, synergistic effects may be observed in patients.

We thank Y. Boucher, D. Buerk, L. Gerweck, A. Hartford, A. Kadambi, W. Ruether, P. Bohlen, and H.D. Suit for their critical comments. We thank S. Roberge, J. K. Kahn, J. Huber, and A. Santiago for their outstanding technical support.