Technical Comments

Response to Comments on “Tumor Response to Radiotherapy Regulated by Endothelial Cell Apoptosis”

Science  12 Dec 2003:
Vol. 302, Issue 5652, pp. 1894e
DOI: 10.1126/science.1091166

Our recent publications (1, 2) have challenged the generally accepted notion that tissue stem cells and clonogens are the sole determinant targets in normal and tumor tissue responses to ionizing radiation, with the host microenvironment providing modulating though not mandatory signals (35). Our genetic and pharmacological studies provided evidence that, at doses comparable to those used in clinical radiotherapy, expression of radiation damage by intestinal crypt or tumor clonogens is conditionally linked to radiation-induced endothelial apoptosis in the affected tissue microvasculature. Microvascular endothelial apoptosis was obligatory: Its abrogation prevented tissue damage altogether.

Suit and Willers (6) argue that we did not provide appropriate discussion of the mechanism of this linkage in (2). We have, however, addressed this issue recently (7). We postulated that at the clinically relevant dose range, radiation-induced microvascular dysfunction couples to the processing of DNA lesions directly induced in irradiated clonogens (7). The microvascular component might involve leakage of a circulating factor, a bystander effect secondary to endothelial damage, or cellular signaling induced by transient local ischemia that converts otherwise repairable DNA lesions into lethal damage. However, the endothelial effect is not limited to modulation of DNA damage repair and postmitotic cell death, but also regulates other radiation-induced tissue dysfunction. For example, inhibition of endothelial apoptosis prevented intestinal crypt shrinkage (7), resulting from S-phase mitotic arrest of crypt epithelial cells coupled with their continued migration toward the villus tip (810). This phenomenon defines endothelium as a distinct primary target, not merely as modulator of post-mitotic death of tissue clonogens.

Our experiments also indicated that at higher doses of 18 to 20 grays (Gy), death of tumor cells (2) or intestinal crypt clonogens (11) becomes independent of the endothelial effect. At these doses, residual DNA damage remaining after the initial phase of DNA double-strand break repair appears sufficient to cause cell death regardless of microvascular dysfunction. Hence, tissue response after escalating doses of radiation consists of two components: microvascular dysfunction-dependent and -independent elements. The demonstration that endothelial damage at the low dose range is fundamental to tumor cure is consistent with the developing literature showing that vascular endothelial growth factor (VEGF) may attenuate, and anti-VEGF strategies may sensitize tumor response to radiation (1214).

Suit and Willers (6) suggest that experiments by Budach et al. (15) indicated that the microvascular response may be irrelevant to the tumor response in vivo. They argue that if radiation sensitivity of the host-derived tumor vasculature mattered, tumors should have been radiosensitive when implanted into severe combined immunodeficient (SCID) mice. This position rests on the assumption that endothelial cells, like other somatic SCID cells, should display radiation hypersensitivity. However, those studies did not include direct assessments of the radiosensitivity of endothelium of tumors growing in SCID mice. We recently carried out such studies and found that endothelial cells purified from B16F1 melanomas growing in SCID/C57BL/6 or wild-type C57BL/6 mice exhibited identical sensitivity to radiation-induced apoptosis (11). Like Budach et al. (15), we found that tumor implantation (MCA/129 sarcoma or B16F1 melanoma) into SCID mice does not affect the tumor response. Only when implanted into mice with genetic alterations impacting endothelial cell apoptosis, such as acid sphingomyelinase–deficient (asmase–/–) or Bax–/– mice, was the tumor response altered, yielding radiation-resistant phenotypes. These data provide a basis for reinterpretation of the SCID TCD responses observed by Budach et al. (15).

Further, the observation that tumor cells arising in SCID mice were 2.6 times more sensitive to radiation than wild-type tumors (16) is not inconsistent with our model. Tumor cells carrying the SCID genotype become so hypersensitive to DNA damage, as demonstrated by the clonogenic assay in vitro (16), that they appear to succumb to postmitotic cell death at doses insufficient for endothelial apoptosis. A similar response pattern has been observed in other cells defective in DNA damage recognition and repair pathways that commonly exhibit extreme hypersensitivity to radiation (17, 18). Our recent studies showed that inactivation of the ataxia-telangiectasia mutated (ATM) gene yields a 2.5-fold increase in sensitivity of C57Bl/6 mice to death from the gastrointestinal (GI) syndrome [100% lethal dose (LD100) of 15 Gy versus 6 Gy in wild-type and ATM–/– mice, respectively (11)], specifically resulting from enhancement of apoptotic and postmitotic death of crypt stem cell clonogens. Endothelial apoptosis, which begins at 8 Gy (1), was unaltered by ATM mutation. Hence the vascular component of the GI response was not engaged in the ATM–/– intestinal response, where direct lethal effects of radiation on crypt clonogens were activated at 4 to 6 Gy.

Both the comment of Suit and Willers (6) and that of Brown et al. (19) note that the 50% tumor control (TCD50) dose of 15 Gy reported by us for MCA/129 fibrosarcomas grown in asmase+/+ mice is particularly low and is virtually unknown for tumor transplants in syngenic hosts. Brown et al. (19) speculate that this could be due to an exceptionally active immune response in these mice. We have, however, carried out extensive histopathologic and immunopathologic studies of the MCA/129 fibrosarcomas and B16F1 melanomas in wild-type and knockout asmase mice, and found that none of the host–tumor combinations displayed the classic criteria of an immunogenic tumor. Hematoxylin and eosin staining detected only rare peritumoral or endotumoral mononuclear, lymphocytic or inflammatory infiltrates, the hallmarks of tumor immunogenicity (20), before or after 15 Gy. The lack of immune cell infiltrates was confirmed using immunohistochemical markers including CD45 (a mononuclear cell common antigen), B220 (a B cell antigen), and CD3 (a T cell receptor). Furthermore, genetic studies showed that tumor growth and radioresponsiveness were unaffected by implantation into rag–/– mice, which lack functional T and B cells (21), or in MEF–/– mice, which lack functional NK and NK-T cells (11, 22). These studies provide definitive evidence that MCA/129 fibrosarcoma and B16F1 melanoma are nonimmunogenic in the model systems tested by us. Finally, Brown et al. (19) avoid the compelling demonstration that the tumor response to radiation was also defective in Bax–/– mice, which do not display a substantive immune response defect (23).

Although the TCD concept is a widely accepted approach to assess curability of tumors with radiation, experience with transplantable tumors in mice has proven this method extremely sensitive to experimental manipulation. Whether the TCD50 for experimental tumors can be predicted by measuring the number of tumor clonogens and their intrinsic radiosensitivity in vitro, as suggested by Suit and Willers (6), is debatable. Data on six experimental tumors studied by Gerweck et al. (24) showed that the measured values of these parameters did not correlate with the TCD50. When figured together into a mathematical model, these parameters provided a rank-order of the TCD50, but the calculated values (approximated range 10 to 22 Gy) consistently underestimated the TCD50 values observed experimentally (30.9 to 86.5 Gy). The gap was decreased to approximately 9 Gy only after adjusting the intrinsic radiosensitivity parameters mathematically to their estimated values under anoxic conditions, and measuring the TCD50 under clamp hypoxia, but the correlations were still imperfect.

We question the legitimacy of these manipulations. Clamp hypoxia is a procedure used by many in TCD experiments to “normalize” for variations in tumor oxygenation (2528). This procedure, however, significantly increases TCD50 values (2528). Experimental data demonstrated that as little as 5 minutes of clamp hypoxia, followed by clamp release and reoxygenation, activates protein kinase C (29), members of all three MAPK families (p38, c-Jun Kinase, ERK) (29, 30), JAK/STATs (31), nitric oxide synthase (32), and NF-κB (33). These pathways engage Cyclin D1 (34), transcribe COX-2 (31), increase the Bcl-2/Bax ratio (33), and upregulate DNA repair enzymes (34, 35). The outcome of these adaptive responses is antiapoptotic protection of endothelium (33), enhanced radioresistance of tumor cells in vitro (3639), and increased TCD50 of experimental tumors (40, 41). Hence, clamp hypoxia appears capable of increasing TCD50 values by both attenuating endothelial apoptosis and improving tumor cell survival from DNA damage.

Another common manipulation that artificially increases TCD50 values is whole-body irradiation, used by many prior to tumor implantation to “inactivate” antitumor host immune systems (26, 28, 42, 43). The increase in TCD50 may result, at least in part, from modifying the function of host bone marrow–derived endothelial precursors required for development of tumor microvasculature, elements sensitive to whole-body irradiation (2, 44, 45). Finally, frequent use of tumor cell lines propagated for many generations in tissue culture may also contribute to the high TCD50 values observed. Such cell lines may represent highly selected clones with inherent resistance to stress, providing a survival advantage under culture conditions, with cross-resistance to radiation.

One group of TCD data free from procedural confounders is the results of single-dose stereotactic radiotherapy used to treat human metastatic tumors to the brain. Multiple recent clinical studies reported 80 to 95% permanent local control of ≤3 cm tumors (that is, from lung, breast, melanoma, renal, colorectal, testicular, gynecological and thyroid tumors) by doses of 18 to 20 Gy, regardless of tumor type (4649). Some of these studies were criticized as underestimating the actual tumor dose because of significant treatment dose inhomogeneities, with various tumor regions receiving more that 50% of the prescribed dose. However, recent data from Memorial Sloan Kettering Cancer Center, using a stereotactic intensity-modulated radiotherapy (IMRT) technique that permits maximal tumor dose inhomogeneities of ±2%, confirmed that metastatic brain tumors are locally cured at rates of 75 to 90% with single doses of 18 to 21 Gy (11). Our observations on the radiosensitivity profiles of MCA/129 fibrosarcoma and B16F1 melanoma are compatible with these clinical data. Whether microvascular damage plays a role in the response of human tumors to radiation is currently being explored.


Navigate This Article