Effects of Overexpression of Dimethylarginine Dimethylaminohydrolase on Tumor Angiogenesis Assessed by Susceptibility Magnetic Resonance Imaging

INTRODUCTION

Angiogenesis is the development of new blood vessels to provide a nutritive blood supply and is a prerequisite for tumor growth and survival (1) . Tumor blood vessels can grow by sprouting, intussusception, or the incorporation of endothelial cell precursors. Angiogenesis is stimulated by the release of specific growth factors from tumor cells, endothelial cells, or associated macrophages, in particular VEGF 4 (2) . Production of these factors is up-regulated by physiological conditions associated with tumors, e.g., hypoxia, low glucose, and pH, all of which are consistent with a compromised blood supply.

Numerous angiogenic stimuli also induce the production of NO, and strong positive correlations between nitric oxide synthase expression and human tumor grade have been demonstrated (3, 4, 5, 6, 7) . Intracellular factors that regulate NO synthesis may therefore represent important targets in the control of tumor growth. We recently showed that C6 glioma cells genetically engineered to constitutively overexpress the enzyme DDAH resulted in tumors that grew twice as fast as the wild type (8) . DDAH metabolizes two competitive inhibitors of NO synthesis, asymmetric dimethylarginine and N-monomethyl-l-arginine, indirectly leading to an increase in NO production. Inhibition of DDAH activity leads to decreased NO production both in vitro and in vivo (9) . Upon excision, tumors derived from C6 cells overexpressing DDAH bled more profusely compared with wild type. Subsequent analysis showed increased expression and secretion of VEGF in tumors overexpressing DDAH (8) .

¹H MRI, with its high temporal and spatial resolution, presents itself as an ideal technique as a noninvasive assay of tumor angiogenesis in vivo (10 , 11) . In particular, two complementary magnetic susceptibility-based MRI approaches have been used in this regard. Firstly, tumor vasculature has been detected using the intrinsic R₂* contrast produced by paramagnetic deoxyhemoglobin within tumor blood vessels, to which gradient-recalled echo MRI sequences are sensitive (12) . This approach, which is truly noninvasive because it relies on endogenous deoxyhemoglobin for contrast, offers an extremely sensitive method of monitoring the dynamics of vascular modeling, function, and regression in vivo (13, 14, 15, 16) . Secondly, tumor blood volume has been assessed using susceptibility contrast MRI measurements, in which the tumor uptake of USPIO particles is measured. These novel blood pool contrast agents act in a similar way to deoxyhemoglobin but are more powerful in creating susceptibility effects that strongly and locally alter the transverse relaxation rates R₂ and R₂* (17 , 18) . Contrary to the conventional contrast agent gadopentetate dimeglumine, these high molecular weight USPIO blood pool contrast agents do not leak as readily from the blood vessels. The resulting long intravascular half-life thus simplifies measurements of tumor R₂ and R₂* during steady-state concentrations of the USPIO particles. We have recently shown that estimates of tumor blood volume, derived from measurements of the absolute changes in the relaxation rates R₂ and R₂* after administration of the USPIO blood pool contrast agent NC100150 (Clariscan; Amersham Health), correlate with the tumor blood volume determined by fluorescence microscopy (19) . Furthermore, the ratio of relaxivities, ΔR₂*/ΔR₂, can be used to derive maps of capillary diameter (19, 20, 21, 22) and has shown good agreement with histologically determined vessel size (20) .

To validate MRI-derived end points, it is essential that appropriate histological analyses of the tumor vasculature are performed (10) . A number of immunohistochemical techniques are available to assess tumor vasculature at a microscopic level, detailed information that is not obtainable by noninvasive, volume-averaging MRI techniques (23, 24) . One such approach is following the tumor uptake of the perfusion marker Hoechst 33342, from which quantitative histologically derived morphometric parameters of tumors have been determined and correlated with noninvasive MRI studies of physiological processes (19 , 25) .

To assess the effects of overexpression of DDAH on tumor angiogenesis, we used both intrinsic susceptibility and susceptibility contrast MRI to assess the tumor vascular architecture and morphology in vivo of tumors derived from C6 glioma cells that constitutively overexpress DDAH and tumors derived from wild-type C6 cells. In addition, fluorescence microscopy of tumor uptake of Hoechst 33342 was quantified to assess the perfused fraction of the tumor.

MATERIALS AND METHODS

Animals and Tumors.

Wild-type C6 cells and clone 27 (D27) C6 glioma cells, transfected with the full coding region of the rat DDAH I gene, were used (8) . Cells (2 × 10⁶) were injected into the flanks of female nu/nu mice under halothane anesthesia. Because tumors derived from mock-transfected cells grew similarly to wild type in vivo (8) , MRI was performed only on C6 wild-type and D27 tumors. Two separate cohorts of each tumor type were used for intrinsic susceptibility and susceptibility contrast MRI. All experiments were performed in accordance with the United Kingdom Home Office Animals Scientific Procedures Act 1986.

MRI.

Size-matched C6 wild-type (mean volume, 1.05 ± 0.2 cm³) and D27 (mean volume, 1.06 ± 0.2 cm³) tumors were imaged at 23 and 18 days postpassage, respectively. Anesthesia was induced with a 10 ml/kg i.p. injection of fentanyl citrate (0.315 mg/ml) plus fluanisone [10 mg/ml (Hypnorm; Janssen Pharmaceutical Ltd.)], midazolam [5 mg/ml (Hypnovel; Roche)], and water (1:1:2). ¹H MRI was performed on a Varian 4.7T instrument using a two-turn 12-mm coil. The anesthetized animals were positioned so the tumor hung vertically into the radiofrequency coil and covered with a warm water blanket to maintain the core temperature at 37°C. Field homogeneity was optimized by shimming on the water signal for each tumor to a linewidth of 30–40 Hz.

Tumor vascular development, maturation, and function were assessed by MGRE MRI (26 , 27) . Images with increasing T₂* weighting were acquired using a gradient echo sequence with TE = 5 to 40 ms, echo spacing of 5 ms, 8 echoes, repetition time = 100 ms, and flip angle α = 45°. Images were acquired from three transverse 1-mm slices through the tumor, using 8 acquisitions of 256 phase encode steps over a 4 × 3-cm field of view for each TE. The total imaging time was ∼9 min. Air, 5% CO₂ in air, or carbogen (5% CO₂/95% O₂) was administered via a nose piece equipped with a scavenger to prevent the leakage of paramagnetic oxygen into the magnet bore, which could potentially change the magnetic susceptibility around the coil and produce image artifacts. MGRE images were first acquired during air breathing, subsequently during 5% CO₂ in air breathing, and finally during carbogen breathing.

Paramagnetic deoxyhemoglobin creates susceptibility perturbations around blood vessels, increasing the magnetic resonance transverse relaxation rate R₂* (R₂* = 1/T₂*, the apparent spin-spin relaxation time). Sampling the MGRE MRI SI at several TEs allows the slope of ln(SI) versus TE to be calculated, which determines R₂* and is directly proportional to the tissue content of deoxyhemoglobin and hence a sensitive index of tissue vascularity. A fast R₂* is consistent with a high [deoxyhemoglobin] and hence indicative of high blood vessel density. In addition, the intercept of the fitted data, i.e., the SI at TE = 0 (A₀), is exclusively sensitive to changes in perfusion. A decrease in the slope (R₂*) implies an increase in oxyhemoglobin. Blood vessel maturation and function were also assessed by measuring the changes in tumor R₂* during hypercapnia (5% CO₂/95% air) as a probe for vascular smooth muscle dilation or during hyperoxia (5% CO₂/95% O₂) to indicate functional blood vessels, respectively (13 , 14 , 16) . Apparent R₂* and A₀ maps were calculated on a pixel-by-pixel basis from the MGRE image data sets (26 , 27) . The average apparent R₂* relaxation rates were calculated for each tumor for a global ROI encompassing the whole tumor image but excluding the surrounding skin and muscle. A relatively fast R₂* is consistent with a high [deoxyhemoglobin] and hence vascularization, but cellular paramagnetic debris from necrosis such as hemosiderin can also increase R₂*. Therefore, the intercept A₀ maps, which are exclusively sensitive to perfusion, were used to identify local ROIs that showed relatively high SI and hence were vascular, and changes in R₂* within them were subsequently assessed (Fig. 1) ⇓ .

Tumor blood volume was assessed by susceptibility contrast MRI measurements, made from a single central transverse 2-mm-thick slice through the tumor (19) . MGRE images were acquired as described above to quantify R₂*. A multispin echo sequence was used to quantify R₂, with repetition time = 300 ms and TE = 11, 20, 30, and 40 ms. Images were acquired from exactly the same 4 × 3-cm field of view. The total imaging time for a complete set of MGRE and multispin echo images was ∼15 min. After acquisition of one set of images, 2.5 mgFe/kg NC100150 [Clariscan; Amersham Health (28 , 29)] was administered via a tail vein, and two additional sets of images were acquired. Apparent R₂ and R₂* maps were calculated on a pixel-by-pixel basis from the image data sets as described above to quantify tumor R₂ and R₂*. Spin echo MRI techniques have a lower sensitivity compared with gradient echo; hence, the increase in R₂ induced by a susceptibility contrast agent is smaller than the increase in R₂*. However, spin echoes are relatively insensitive to deoxyhemoglobin-induced macroscopic susceptibility gradients and erythrocyte blood flow (which are essentially artifactual), to which gradient echoes are highly sensitive. Therefore, changes in R₂ reflect the capillary microvasculature, making them useful in depicting functional/physiological vascular changes, e.g., angiogenesis, whereas R₂* is sensitive to both micro- and macrovasculature (17 , 18 , 20) . Tumor vascular morphology was also assessed by first deriving maps of ΔR₂* and ΔR₂, the differences between the relaxation rates before and after injection of NC100150. From these, the ratio map of ΔR₂*/ΔR₂ was derived. The ratio ΔR₂*/ΔR₂ increases with increasing vessel size and was thus used as an index of the average microvessel size over the whole tumor (19 , 20 , 22) .

Fluorescence Microscopy of a Perfusion Marker.

Tumor-bearing mice received i.v. injection via a tail vein of 15 mg/kg of the perfusion marker Hoechst 33342 in PBS (19 , 23 , 24) . Hoechst 33342 stains the nuclei of endothelial cells and cells adjacent to tumor blood vessels that are perfused at the time of i.v. injection and hence delineates functional tumor vasculature. One min after injection, the animals were killed by cervical dislocation to prevent the Hoechst dye from diffusing too far from perfused vessels, and the tumors were excised rapidly, frozen, and stored in liquid nitrogen. Frozen sections (10 μm) were subsequently cut. Fluorescence signals of whole tumor sections were recorded using a motorized scanning stage (Marzhauser) driven by Stage-Pro (Media Cybernetics) on an IX70 fluorescence microscope (Olympus Optical Co. Ltd.). Digital images from both C6 and D27 tumors were acquired using the same exposure time, and composite images were then synthesized from each field using Image-Pro Plus software (Media Cybernetics). Fluorescent particles were detected above a threshold that was constant for all of the composite images (analySIS; Soft Imaging System), and the area of the tumor section with Hoechst 33342 fluorescence was then determined and expressed as a percentage of the whole tumor section (mean perfused area). Because the images were acquired and analyzed under identical conditions, any differences in the mean perfused area would result from differences in tumor perfusion.

RESULTS

Synthesized A₀ and R₂* maps obtained from one C6 wild-type and one D27 tumor while the host breathed air, 5% CO₂/95% air, and, finally, 5% CO₂/95% O₂ are shown in Fig. 2 ⇓ . Overall, both D27 and wild-type tumors exhibited heterogeneity in the R₂* maps, with regions of both high and low SI identified within the tumor. Tumors derived from D27-transfected cells were more vascularized compared with tumors derived from wild-type cells, as indicated by the significantly faster average baseline R₂* shown in Fig. 3 ⇓ . No significant change in tumor R₂* over the whole tumor could be detected in either of the tumor lines in response to 5% CO₂/95% air or 5% CO₂/95% O₂. In an approach analogous to histological vascular hot spot analysis, local ROIs showing relatively high SI in the perfusion-sensitive A₀ intercept maps were identified and were predominantly found within the rim of the tumor (Figs. 1 ⇓ and 2 ⇓ ). The baseline R₂* measured from these local ROIs was also significantly faster in the D27 tumors compared with C6 wild type. Small decreases in R₂* were measured from these vascular hot spots in both tumor lines during hypercapnia and hyperoxia (Fig. 3) ⇓ .

Synthesized R₂ and R₂* maps from one C6 wild-type tumor and one D27 tumor before administration of 2.5 mgFe/kg NC100150 and the ΔR₂ and ΔR₂* maps acquired in the presence of the USPIO particles are shown in Fig. 4 ⇓ . Again, both C6 wild-type and D27 tumors exhibited heterogeneity in the R₂* maps both before and after administration of the blood pool agent, whereas the R₂ maps were more homogeneous. The baseline R₂* and R₂ and the changes induced by NC100150 are shown in Fig. 5 ⇓ . There was no significant difference between the average R₂* of the two cohorts of each tumor type used in this study and those in the MGRE MRI experiments. As before, the average baseline R₂* measured over the whole tumor was significantly faster in the D27 tumors than in the C6 wild-type tumors. NC100150 induced significant increases in both R₂* and R₂ of both C6 wild-type and D27 tumors, and this response was significantly greater in the D27 tumors. This response was sustained in the second set of images acquired after administration of NC100150, suggesting negligible leakage of NC100150 out of the tumor vasculature over the 30-min time period (data not shown).

Tumor vascular morphology was also assessed by deriving maps of the ΔR₂*/ΔR₂ ratio, a ratio that increases with increasing microvessel size. Fig. 6 ⇓ shows calculated ΔR₂*/ΔR₂ maps of the same tumors in Fig. 4 ⇓ . The intensity of the synthesized ΔR₂*/ΔR₂ maps of both C6 and D27 tumors is similar. The average ΔR₂*/ΔR₂ ratio for both C6 and D27 tumors is also summarized in Fig. 6c ⇓ , and there was no significant difference between the two tumor types.

Analysis of the tumor perfusion was performed on sections obtained from mice treated with Hoechst 33342. Fig. 7 ⇓ shows composite fluorescence microscopy images of the perfused vascular architecture obtained from one C6 wild-type tumor and one D27 glioma. A greater abundance of Hoechst 33342 staining was associated with the D27 tumors, consistent with the D27 tumors having more perfused vessels. The mean perfused areas obtained for the two tumor types are summarized in Fig. 7c ⇓ , which shows that the perfusion of the D27 tumors was significantly greater than that of the C6 wild-type tumors.

DISCUSSION

Investigation of tumor angiogenesis in vivo within genetically manipulated tumors expressing a well-defined phenotypic change gives an opportunity to investigate the role of factors implicated in tumor vascular morphogenesis in vivo and the chance to identify, validate, and calibrate MRI-derived prognostic and diagnostic indices that can be translated into the clinic (13 , 14 , 16 , 30) . We have used this powerful approach to evaluate the effects of DDAH on tumor angiogenesis in vivo.

MGRE MRI permits efficient, quantitative assessment of the role of factors implicated in tumor blood vessel development. Image contrast arises only from intravascular RBCs, and contrast changes are not sensitive to vascular permeability. The significantly faster baseline R₂* of the D27 tumors is consistent with these more rapidly growing D27 gliomas having a greater vascular development (i.e., angiogenesis) compared with wild-type C6 gliomas (8) .

The maturation state of tumor vasculature has recently been suggested as a marker of tumor progression, with the recruitment of pericytes resulting in tumor blood vessels refractory to withdrawal of vascular growth factors such as VEGF (13 , 14) . Clinical measures of capillary maturation may thus assist the identification of patients who would benefit from antiangiogenic therapies or vascular targeting treatments (31, 32, 33) . In this study, the maturation and functional state of the tumor blood vessels were probed by measuring the ΔR₂* in response to hypercapnia and hyperoxia, respectively. Small decreases in R₂* were identified within the periphery of tumors derived from both wild-type C6 and D27 cells, presumably indicating areas of active angiogenesis. These results are consistent with the blood vessels of both tumor types containing (a) some smooth muscle in the vessel walls and (b) deoxygenated blood. There was no significant difference in either response between the C6 and D27 gliomas.

Administration of 2.5 mgFe/kg NC100150 was a sufficient dose at 4.7 Tesla to induce significant changes in the transverse relaxation rates R₂* and R₂ of both wild-type C6 and D27 gliomas, and this response was heterogeneous over the whole tumor. There was no apparent recovery to baseline values of either R₂* and R₂ within the 30 min after the administration of NC100150, implying negligible permeability of NC100150 from the tumor circulation. The significantly larger changes in R₂* and R₂ measured in the D27 tumors are consistent with a larger blood volume compared with C6 wild type. This was supported by tumor uptake of Hoechst 33342, which showed that the perfused area in the D27 tumors was significantly greater than that in the C6 wild-type tumors (Fig. 7) ⇓ and confirmed our earlier observations (8) . These data highlight the advantages of using histological techniques such as Hoechst 33342 that assay for perfused functional vasculature. These methods appear to correlate with angiogenic potential, whereas with blood vessel density measurements using pan-endothelial markers, the correlation with angiogenesis can be unclear (34) . The increased blood volume of the D27 tumors is consistent with clinical observations, in which high-grade human gliomas had a substantially higher relative blood volume compared with low-grade tumors (35) .

Evaluation of the ΔR₂*/ΔR₂ ratio, an index of the average microvessel size, showed no significant differences between the D27 and wild-type C6 gliomas (Fig. 6) ⇓ . This implies that, despite the clearly increased tumor blood volume in the D27 tumors, the caliber of the vasculature in the two tumor types is similar. Using the methodology of Tropres et al. (22) , we have recently shown that estimates of both the tumor fractional blood volume and microvessel size could be calculated using the NC100150-induced changes in R₂* and the ΔR₂*/ΔR₂ ratio, respectively (19) . Assuming that the susceptibility change induced by NC100150 was the same for both tumor types, and using a previously reported diffusion coefficient for C6 gliomas of 1.1 × 10⁻⁹ m²s⁻¹ (36) , the fractional blood volume of the D27 and C6 gliomas was estimated to be 3% and 1%, respectively, whereas the average microvessel size for both tumor types was 6 μm. This is not dissimilar from previous histological measurements (mean ± 1 SE) of the microvessel diameter of C6 gliomas of 7.8 ± 0.5 μm (37) and 12.5 ± 3.9 μm (20) .

Taken together, the results are consistent with enhanced tumor angiogenesis in D27 tumors and suggest that DDAH activity might play a central role in tumor growth by regulating the concentration of tumor-derived NO. The spatial and temporal production of NO is tightly regulated and could be an important factor in determining tumor progression. We have previously shown enhanced VEGF production by D27 tumors compared with C6 wild-type tumors, suggesting that DDAH may enhance tumor growth and angiogenesis via VEGF expression (8) . The lack of significant difference in (a) the hypercapnic/hyperoxic response and (b) vessel caliber between the C6 and D27 tumors suggests that DDAH is primarily involved in the initial formation of tumor blood vessels, either through angiogenic sprouting or intussusception, rather than the subsequent stages of vascular remodeling (1) . Support for this has come from in vitro invasion assays reported previously (8) .

Intrinsic susceptibility-based MRI measurements were made on a tumor system in which DDAH was constitutively overexpressed, and the results were compared with wild type. A similar approach has been used previously to investigate the role of hypoxia-inducible factor 1α in tumor angiogenesis (38) . The localized tumor ΔR₂* responses to hypercapnia and hyperoxia measured by MGRE MRI herein were inherently small, a consequence of the reduced contrast:noise ratio of intrinsic contrast-enhanced MRI. Murine erythrocytes in whole blood are typically 6 μm in diameter (19) . Thus, the ability of murine erythrocytes, the primary source of changes in R₂* intrinsic contrast, to traverse similarly sized, tortuous capillaries in vivo would be limited, abrogating the hypercapnic/hyperoxic response (19) . This would explain why, contrary to the hyperoxic ΔR₂* response, a higher proportion of functional vasculature was measured in the D27 tumors after Hoechst 33342 uptake. Susceptibility contrast MRI has been used previously to show an increased blood volume fraction in murine MCF-7 mammary carcinomas constitutively overexpressing VEGF compared with nontransfected controls (30) . In this study, we have clearly demonstrated the utility of NC100150 to show that overexpression of DDAH in C6 gliomas resulted in a greater tumor blood volume compared with wild type, yet the sizes of the vessels in the two tumor types were similar. An alternative approach to investigate both the effects and dynamics of DDAH on vascular remodeling assessed by intrinsic susceptibility MRI and the contribution of DDAH to vascular morphology assessed by the microvessel size index ratio ΔR₂*/ΔR₂ after susceptibility contrast MRI would be to use a tumor system in which DDAH is under inducible control, rather than being constitutively overexpressed, and where each tumor acts as its own control. A C6 cell line in which production of DDAH is under the control of a tetracycline-inducible promoter is currently being engineered.

The development and validation of quantitative, clinically applicable MRI end points of functional/physiological measures of tumor angiogenesis is critical for (a) enhancing our understanding of the incipient tumor vasculature and (b) determining the efficacy of antiangiogenic and antivascular therapies, many of which do not induce a significant growth delay in human tumor xenografts (39) . The utility of both MRI methods used herein in the clinic has been demonstrated. Recently, MGRE MRI, which can be implemented on most clinical MRI scanners, was used to measure human tumor R₂* during air and carbogen breathing to assess the prognostic value of ΔR₂* for radiotherapeutic outcome (40) . The blood pool agent NC100150 was used recently to measure vessel permeability and blood volume in human breast cancers (41) . The methods also present themselves as simpler than dynamic T₁-weighted contrast-enhanced MRI using gadopentetate dimeglumine approaches to employ in the clinic because there is no need for rapid data acquisition, and the data analysis and interpretation are more facile. In particular, noninvasive susceptibility-based MRI incorporating USPIO contrast agents such as NC100150 appears very promising.

In summary, both baseline intrinsic susceptibility (deoxyhemoglobin) and steady-state contrast-enhanced (NC100150) MRI measurements were consistent with increased tumor vascular development and blood volume in the D27 tumors compared to wild-type C6 gliomas. The results are consistent with DDAH having an important role in tumor growth and angiogenesis, particularly in the initial stages of vasculogenesis.