1 Introduction

It is well established that tumor vasculature plays a key role in tumor growth [1]. Tumor vessels are not only morphologically, but also physiologically distinct from host vessels [2]. Knowledge about these aspects of tumor vasculature is critical to tumor therapy, both in novel anti-angiogenesis therapeutic treatments and in conventional cytotoxic therapy. Moreover, in recent studies, blood hemodynamic modifiers have been employed to enhance the effects of treatment [3]. However, the exact mechanism by which blood hemodynamic modifiers act is still not well understood [4] and their effects on tumor hemodynamics remain unclear [5]. It is crucial, therefore, to characterize the tumor vasculature and its response to blood hemodynamic modifiers in order to assess and improve the effectiveness of current therapeutic treatments.

Magnetic resonance imaging (MRI), in combination with contrast agent administration, has proven useful in the characterization of tumor vasculature. Several studies have shown that magnetic susceptibility effects, caused by blood-pool contrast agents that consist of ultrasmall superparamagnetic particles of iron oxide (USPIO), can be used to assess blood volume and vessel sizes within tumors. USPIOs have been mostly employed in combination with quantitative and/or qualitative T 2 and T 2* measurements [6–12]. In particular, the enhancement in the transverse relaxation rates R 2 (=1/T 2) and R 2* (=1/T 2*), after administration of USPIO, provides an index proportional to the blood volume of the microvasculature and macrovasculature, respectively. However, quantitative ΔR 2 and ΔR 2* measurements are characterized by relatively long acquisition times, due to the long repetition time TR needed to minimize T 1 effects.

Quantitative T 1 MRI, in combination with USPIO administration, represents an alternative approach to T 2/T 2* USPIO contrast-enhanced MRI, in particular, when fast assessment (i.e., on the order of seconds) of blood hemodynamics is needed. In this study, detailed measurements of changes in longitudinal relaxation rate R 1 (=1/T 1), following USPIO administration were performed within in vivo tumor tissue. R 1 was measured with the inversion–recovery snapshot fast low-angle shot (IR-FLASH) imaging sequence pre- and post-USPIO administration, in two different types of murine colon carcinoma (C26a, C38). The findings were compared with the vascularity of these tumor types as determined by immunohistochemistry. To assess whether USPIO-induced R 1 changes provide a means to assess blood hemodynamics with high temporal resolution, we administered hydralazine—a central vasodilator which induces vasoconstriction in subcutaneous tumors by the steal effect mechanism [13]—and monitored R 1 changes in time.

2 Materials and Methods

2.1 MRI

All procedures were approved by the Radboud University Nijmegen Medical Center Animal Care and Use Committee. C26a (n = 9) and C38 (n = 9) murine colon tumor tissue fragments were implanted subcutaneously in female Balb/C and C57/Bl6 mice, respectively, of 8–12 weeks of age. Experiments were performed when tumors reached the diameter of approximately 0.8 cm. Mice were anesthetized with isoflurane inhalation (1.5–2%); in each mouse, a catheter was inserted in the tail vein for the administration of the contrast agent and the tumor was positioned in the center of a 10-mm diameter surface radio-frequency (RF) coil used as a transmitter/receiver. During the measurements, the body temperature was monitored with a rectal fluoroptic probe (Luxtron 712, Luxtron Corporation, California, USA) and maintained at a constant temperature of 37 ± 1°C with a warm water pad.

MRI experiments were performed on a 7T/200 mm horizontal-bore MR spectrometer interfaced to a SMIS console and equipped with a gradient insert with a gradient strength of 150 mT/m and rise time of 150 μs. The image acquisition protocol started with three-gradient echo scout images followed by multislice gradient echo images for anatomical localization of the tumor. Imaging parameters were: repetition time (TR) = 400 ms, echo time (TE) = 10 ms, image matrix size of 128 × 128, field of view (FOV) of 5.8 cm × 5.8 cm, slice thickness (SLT) of 0.7 mm and 1 excitation per phase-encoding step. IR-FLASH imaging [14, 15] was then performed on a slice through the center of the tumor, prior to and following administration of an USPIO blood-pool contrast agent (Sinerem®, Guerbet, France; 150 μmol Fe/kg). Imaging parameters were: TR/TE = 5 ms/2.7 ms, image matrix size of 64 × 64, FOV of 3 cm × 3 cm, SLT of 1.6 mm. In this sequence, the longitudinal magnetization is first inverted by a hyperbolic secant adiabatic inversion pulse and then repeatedly sampled by a train of small flip-angle (5°) read-out Gaussian pulses to generate multiple images at different time points on the T 1 water proton recovery curve. Digitized RF spoiling pulses with phase angle increments of 117° [16], and a gradient spoiling were applied in order to spoil the transverse steady-state magnetization prior to each read-out pulse. The time interval between the inversion pulse and the first image was 52 ms. Ten dummy scans were applied prior to each image acquisition so that the total acquisition time per image was 370 ms. The lines in the k-space were acquired with a centric profile order to generate 16 images, which sampled the recovery of the longitudinal magnetization over ca. 5 s, at intervals of 370 ms.

To assess the precision of the R 1 measurement, in order to determine whether the changes in R 1 induced by the contrast agent are greater than the measurement error, four repeated R 1 measurements were performed prior to and following administration of USPIO on a group of mice (n = 5, C26a tumors and n = 5, C38 tumors). Further, in a group of mice with C38 tumors (n = 3), an intraperitoneal catheter was inserted for hydralazine injection. Hydralazine was dissolved in NaCl 0.9% to a concentration of 0.33 mg/ml and intraperitoneally injected at a dose of 2.5 mg/kg of mouse weight. The MRI protocol was as before with the only difference that after the post-USPIO R 1 measurement, hydralazine was injected, and the R 1 value was measured 5 min thereafter.

2.2 Data Analysis

The Levenberg–Marquardt non-linear least squares algorithm was used to analyze the R 1 relaxation data. R 1 maps were obtained from the entire set of 16 images. Voxel-by-voxel R 1 maps, pre- and post-USPIO, and after hydralazine, were generated from a three-parameter fit of the image intensities according to the equation: S(TI) = A + B exp(−TI/T 1*) and the value of the corrected R 1 was calculated from the formula: R 1 = (T 1*(B/A − 1))−1 which is valid in the small flip-angle limit. Pixel-by-pixel T 1 maps were calculated using an algorithm described elsewhere [17]. First, for each pixel, the signal intensity of the 16 magnitude images was evaluated and the minimum value (S min) was determined. Second, an inversion–recovery curve (i.e., with negative and positive values) was generated by inverting all the data on the left of the minimum value S min. The fit was determined, as well as the coefficient of determination (R 2), which is a measure of the goodness of the fit. Third, the same procedure was repeated choosing one point on the left (S min−1) and one point on the right (S min+1) of S min, so that two additional inversion–recovery curves were generated by inverting all the data on the left of (S min−1) and of (S min+1), respectively. The fit was determined, as well as R 2. Thus, in total, three inversion–recovery curves were generated. The inversion–recovery which displayed the best fit (i.e., highest R 2) was chosen for calculating the T 1 relaxation time. The fitting algorithm is further described in Ref. [17]. From the voxel-by-voxel R 1 maps, ΔR 1 maps (ΔR 1 = R 1post − R 1pre, where R 1post and R 1pre are the R 1 post- and pre-USPIO, respectively) were calculated. For each tumor, the mean ΔR 1 was then calculated by drawing a region-of-interest (ROI)—which included the tumor core—on the ΔR 1 map and averaging the ΔR 1 values of all voxels within the ROI.

To assess the precision of the measured R 1 values, we performed four repeated measurements of R 1 prior to and following administration of USPIO. For each tumor, four R 1 maps pre-USPIO and four R 1 maps post-USPIO were generated as described in the previous paragraph. From each R 1 map, the mean R 1 was calculated by averaging the R 1 values of all voxels within an ROI which included the tumor core. The standard deviation (SD) over four repeated R 1 measurements, pre-USPIO and post-USPIO was calculated [18]. All data are expressed as mean ± SD. The Levenberg–Marquardt non-linear least squares algorithms were implemented in MatLab (Mathworks, Natick, MA, USA).

In order to compare the changes in the R 1 relaxation time (ΔR 1) in the C26a tumor with the ΔR 1 value in the C38 tumor line, the data were further analyzed with the software package Prism (GraphPad Software Inc., CA, USA). The statistical significance (P < 0.001) of differences between the ΔR 1 value in the C26a and C38 tumors was determined by means of a two-tailed Student’s t test.

2.3 Theory

Since the USPIO remains intravascular, modeling of the effects of the USPIO on the R 1 relaxation rate requires the framework of a two-compartment exchange model, where the two compartments are the intra- and extravascular space. The relaxation behavior of the two-compartment exchange model depends on the exchange rate τ−1, defined as the sum of \( \tau_{\text{i}}^{ - 1} \) and \( \tau_{\text{e}}^{ - 1} \), where τ i and τ e are the water proton lifetimes in the intra- and extravascular compartment, respectively, and on the absolute value of the difference between the water proton relaxation rates of the intra- and extravascular space (R 1i and R 1e, respectively). The precise value of τ −1 is not exactly known [19]; however, the reported values are ca. 1 Hz [20–23]. The two-compartment model is in slow exchange when the condition τ −1 ≪ |R 1i − R 1e| is satisfied. Schwarzbauer et al. [24] showed that, following contrast agent administration, in the case of slow exchange the observed change in relaxation rate ΔR 1 is equal to PS/(λ − RBV), where PS is the permeability-surface area product for water across blood vessel walls, λ is the tissue-to-blood partition coefficient and RBV is the regional blood volume. For typical values of λ and RBV [24], RBV ≪ λ and therefore, after Taylor series expansion (1/[1 − x] ~ [1 + x], for |x| ≪ 1), ΔR 1 ~ (PS/λ)(1 + RBV/λ).

2.4 Immunohistochemistry

In two separate groups of mice (n = 6 for C38 and n = 5 for C26a) with tumors of approximately the same size as used for the MR experiments, tumor vascularity was determined by immunohistochemistry. After the animals were killed by cervical dislocation, their tumors were excised immediately and stored under liquid nitrogen. Frozen tumor sections of 5 μm thickness were cut for staining and further analysis of vasculature. After thawing, the sections were fixed in cold (4°C) acetone for 10 min. Between all consecutive steps of the staining procedure, sections were rinsed three times for 2 min in phosphate buffered saline. Sections were mounted in Fluorostab (Organon, Boxtel, The Netherlands). Endothelial structures were stained with 9F1, which is a rat monoclonal antibody to mouse endothelium (Department of Pathology, University Medical Centre Nijmegen, The Netherlands) [25]. Then sections were incubated for 30 min at 37°C with goat anti-rat-Alexa546 (Molecular Probes, Leiden, The Netherlands) and diluted 1:200 in polyclonal liquid diluent (Euro-DPC, Breda, The Netherlands). Quantitative data for tumor vascularity were acquired by a semiautomatic method based on a computerized digital image analysis system, as described previously [26, 27]. Whole tumor sections were scanned for 9F1 positive structures and a contour line was drawn to delineate the tumor area thereby excluding non-tumor tissue from the analyses. Consecutive hematoxylin and eosin-stained tumor sections were employed to distinguish parenchyma from non-tumor tissue. The resulting composite image was divided into smaller ROIs (230 × 230 μm). In each ROI the relative vascular area (RVA: 9F1 positive area divided by the viable tumor area) was calculated. From this, a new image map was created in which the voxels had values of the RVA values in the corresponding ROIs.

3 Results

The IR-FLASH provided a fast method for measuring the water proton R 1 relaxation rate in murine colon carcinoma. Figure 1 shows the sixteen IR-FLASH images of a central slice through a C26a murine colon carcinoma. The acquisition time for the whole set of images was ca. 5 s. The combination of RF and gradient spoiling implemented in the IR-FLASH measurements provided images virtually free from artifacts. The signal intensity of the last image (image 16) was typically less than the signal intensity of the first image, due to small saturating effect originating from the read-out pulses. At the inversion time of approximately 1,000 ms, the longitudinal magnetization goes through the null point. The signal intensity changes from a single voxel within the tumor, as well as the fit of the signal recovery to a monoexponential function are illustrated in Fig. 1 (right). The R 2 value of 0.9996 indicates the excellent agreement between data points and fit.

Fig. 1
figure 1

IR-FLASH measurements of the R 1 relaxation rate in a C26a murine colon carcinoma. The sixteen IR-FLASH images of a C26a murine colon carcinoma (left). Plot of signal amplitudes from a single voxel within the tumor (right). The dashed line indicates the monoexponential fit (R 2 = 0.9996)

Full size image

To estimate the precision of the measured R 1 values of each tumor, four repeated measurements were performed prior to and following USPIO administration (Fig. 2). The column graph shows the values of R 1 obtained from the four repeated measurements prior to and following administration of USPIO, in a C26a and a C38 tumor. The horizontal line indicates the mean of the four measurements, while each point represents the value of one measurement. Prior to USPIO administration, R 1 = 0.5467 ± 0.0024 s−1 in the C26a and R 1 = 0.5791 ± 0.0029 s−1 in the C38. Following USPIO administration, R 1 = 0.5692 ± 0.0013 s−1 in the C26a and R 1 = 0.6596 ± 0.0030 s−1 in the C38. As a result, in both tumors, the USPIO-induced change ΔR 1 was greater than the SD value. The repeated measurements were performed in ten tumors. The results of the data analysis for all tumors are given in Table 1. The SD pre- and post-USPIO administration was typically ca. 1–3 10−3 s−1. A small increase in SD in the post-USPIO measurements was observed. On the other hand, changes in R 1 induced by USPIO (third row in Table 1) were well above SD values in all cases.

Fig. 2
figure 2

Column graph of the changes in the R 1 relaxation rate following USPIO administration in a C26a and a C38 colon carcinoma. Symbols indicate the values of R 1 from four repeated measurements performed prior to and following USPIO administration. Prior to USPIO administration, in the example shown here, R 1 = 0.5467 ± 0.0024 s−1 in the C26a and R 1 = 0.5791 ± 0.0029 s−1 in the C38. Following USPIO administration, R 1 = 0.5692 ± 0.0013 s−1 in the C26a and R 1 = 0.6596 ± 0.0030 s−1 in the C38. In both tumors, the post-USPIO R 1 is significantly higher than pre-USPIO R 1

Full size image
Table 1 SD of R 1 calculated from four repeated measurements, performed in ten tumors and the change in R 1 induced by the USPIO
Full size table

In Fig. 3, a column graph of the changes in the R 1 relaxation rate following USPIO administration in all tumors showed a clear difference between C38 and C26a tumors. The R 1 increase in the C38 tumors (ΔR 1 = 0.072 ± 0.0081 s−1) was significantly larger than in the C26a tumors (ΔR 1 = 0.032 ± 0.0018 s−1, P < 0.001). Among the C26a tumors, changes in R 1 were very similar, while the C38 tumors displayed a broader range of changes in R 1. It should be noted that the smallest ΔR 1 observed in the C38 group was greater than the largest ΔR 1 of the C26a group.

Fig. 3
figure 3

Column graph of the changes in water proton R 1 relaxation rate (ΔR 1) following USPIO administration in the C26a and C38 colon carcinoma. The R 1 increase in the C38 (ΔR 1 = 0.072 ± 0.0081 s−1) was significantly larger than in the C26a colon carcinoma (ΔR 1 = 0.032 ± 0.0018 s−1, P < 0.001). The ΔR 1 in all C38 colon carcinoma was larger than 0.045 s−1 (dotted line), while the ΔR 1 in all C26a colon carcinoma was smaller than this value

Full size image

The immunohistochemically determined RVA of the C38 tumors was larger than the RVA of the C26a tumors (0.059 ± 0.015 vs. 0.020 ± 0.011; P = 0.0013). This is illustrated in Fig. 4, which shows an RVA map, as well as the vascular structure, of a typical C38 and C26a tumor. The photomicrograph clearly shows the substantial difference in vascular volume and morphology between the C26a and C38 tumors.

Fig. 4
figure 4

RVA maps of the C26a (a) and C38 (b) tumor section. Each pixel represents a 0.23 × 0.23 mm ROI. Maps are scaled to 100% which corresponds to an RVA value of 21%. c and d Detailed photomicrographs of the tumors in a and b, showing the typical vascular structure of the C26a (c) and the C38 (d) colon carcinoma

Full size image

Figure 5 shows the R 1 relaxation rate before USPIO administration and changes in the R 1 relaxation rate after USPIO and hydralazine administration. The increase in R 1 after USPIO administration was then followed by a decrease in R 1, after hydralazine administration.

Fig. 5
figure 5

Graph of normalized R 1 prior to USPIO administration (first time point), following USPIO administration (second time point) and 5 min after hydralazine administration (third time point) in three C38 colon carcinomas. Following hydralazine administration, R 1 decreased in all tumors, indicating a vasoconstriction effect

Full size image

4 Discussion

In the current study, we propose a method which uses fast R 1 measurements on tumors prior to and following USPIO administration in order to monitor the tumor hemodynamics. While USPIO-induced changes in the transverse relaxation rates R 2 and R 2* within tumors have been experimentally investigated in a number of studies, less attention has been paid to quantitative measurements of R 1 changes [28]. Here, we first performed detailed measurements of R 1, to assess the precision of our experimental setup, in order to evaluate the significance of the R 1 changes. The high signal-to-noise ratio, achieved by (i) the use of a small surface coil, (ii) the positioning of the imaged slice in the plane of the coil itself and (iii) the high field strength, results in a reliable and precise measure of R 1. As a consequence, R 1 changes following USPIO are significantly larger than the experimental error associated with the R 1 measurement.

The USPIO-induced increase in R 1 may be attributed to the exchange of water protons between the intra- and extravascular space [19]. Following administration of USPIO, R 1 of the intravascular water protons in blood significantly increases (i.e., T 1 decreases from a value of ca. 2 s to a value below 200 ms [28]), whereas the extravascular water protons affected by the contrast agent are only those which are in exchange with the intravascular space. As a result, following USPIO administration, the value of |R 1i − R 1e| increases to a value of ca. 5 Hz, which is greater than the suggested values of τ (~1 Hz [20–23]). Thus, USPIO administration drives the system towards slow exchange. In the limit of slow exchange, two decaying components are expected. However, since only about 5% of the water signal can be attributed to intravascular water protons and since the T 1 relaxation time of the intravascular component is very short, MRI using the inversion–recovery imaging sequences cannot resolve the two decaying components, that is, it cannot detect the fast-decaying component. The measured T 1 is therefore representative of the extravascular component. In the case of slow exchange, as pointed out in Sect. 2, ΔR 1 is proportional to the blood volume and to the PS product. Therefore, differences in ΔR 1 between various tumor lines may be due to differences in blood volume, PS product or both. In our study, there was a statistically significant difference (P < 0.001) between ΔR 1 in the C26a and C38 colon carcinoma, with ΔR 1 being larger in the C38 than in the C26a colon carcinoma. This result indicates a difference in tumor vascular morphology between the two tumor lines, which is attributed either to differences in the PS product or in blood volume. The results from the immunohistochemistry, which was performed in the current study on the same tumor lines, indicate that there was a difference in RVA between the two tumor lines, with the C38 colon carcinoma exhibiting a higher RVA.

More relevant and of immediate interpretation are the R 1 changes for investigating hemodynamic effects in tumors. In fact, in case of vasoconstriction or vasodilation within the same tumor, ΔR 1 can be strictly related to vasoconstriction or vasodilation effects. In other words, if—as a response to a blood hemodynamic modifier—vasoconstriction (or vasodilation) occurs within the tumor, there will be a decrease (or increase) of both surface area and blood volume. Therefore, any vasoconstriction (or vasodilation) is reflected in a decrease (or increase) of R 1. The decrease in R 1 after hydralazine administration, which is a well-known central vasodilator, suggested a hydralazine-induced vasoconstriction in tumors. This is in line with the observation of vasoconstriction by the steal effect mechanism [29] due to hydralazine reported in other studies of subcutaneous tumors utilizing fluorescent staining [30] and contrast-enhanced MRI [31]. The ‘paradoxical’ vasoconstrictor effect (the ‘steal effect’) in tumor vasculature, as a result of administration of vasodilator agents, such as hydralazine for instance [29–31], can be explained in the following way: the systemic vasodilation of all ‘healthy’ blood vessels in the body takes away blood (“steal” the blood) from the tumor vasculature. In fact, since the tumor vasculature itself does not respond directly to the vasoactive agent, as it has a minimal capability of blood flow autoregulation [13], a reduction in tumor blood flow/volume is observed. Tumors have developed a number of ways to ensure a proper blood supply, such as incorporation of pre-existent vessels (also referred to as vessel co-option), vessel modulation (e.g., dilatation and intussusception) and angiogenesis [1, 8]. The ability to monitor tumor blood volume with a high temporal resolution in vivo will provide essential insight into tumor physiology and is a prerequisite to evaluate tumor response to therapy. It will further improve our knowledge of a tumor vasculature response to vasoactive agents (such as in serial measurements of CO2/O2 breathing experiments, for instance), tumor capability for blood flow autoregulation, and transient hemodynamic effects in tumors.

The most significant advantage of quantitative R 1 over quantitative R 2*/R 2 measurements is the temporal resolution: a few seconds for R 1 measurements versus minutes in R 2 measurements. It should be noted that a temporal resolution of seconds could be also achieved with T 2*/T 2-weighted imaging; on the other hand, measurement of relative changes in the signal intensity makes interindividual comparison less reliable. With respect to fast quantitative T 2*/T 2 measurements, based on a steady-state free precession type of acquisition [32], for instance, they suffer from B 1 inhomogeneties and artifacts due to off-resonance (B 0 inhomogeneities), in particular, at high fields. Quantitative R 1 measurements could be an interesting alternative to T 2*/T 2-weighted images and quantitative R 2*/R 2 measurements in assessing changes in blood hemodynamics and thus can give novel insight into the effects of blood hemodynamic modifiers on tumor vasculature, non-invasively and with a high temporal resolution.