Multimodal imaging enables early detection and characterization of changes in tumor permeability of brain metastases

Abstract

Our goal was to develop strategies to quantify the accumulation of model therapeutics in small brain metastases using multimodal imaging, in order to enhance the potential for successful treatment. Human melanoma cells were injected into the left cardiac ventricle of immunodeficient mice. Bioluminescent, MR and PET imaging were applied to evaluate the limits of detection and potential for contrast agent extravasation in small brain metastases. A pharmacokinetic model was applied to estimate vascular permeability. Bioluminescent imaging after injecting D-Luciferin (molecular weight (MW) 320D) suggested tumor cell extravasation had already occurred at week 1, which was confirmed by histology. 7T T1w MRI at week 4 was able to detect non-leaky 100 μm sized lesions and leaky tumors with diameters down to 200 μm after contrast injection at week 5. PET imaging showed that ¹⁸F-FLT (MW 244D) accumulated in the brain at week 4. Gadolinium-based MRI tracers (MW 559D and 2.066kD) extravasated after 5 weeks (tumor diameter 600 μm), and the lower MW agent cleared more rapidly from the tumor (mean apparent permeabilities 2.27×10^-5 cm/s versus 1.12×10^-5 cm/s). PET imaging further demonstrated tumor permeability to ⁶⁴Cu-BSA (MW 65.55kD) at week 6 (tumor diameter 700 μm). In conclusion, high field T1w MRI without contrast may improve the detection limit of small brain metastases, allowing for earlier diagnosis of patients, although the smallest lesions detected with T1w MRI were permeable only to D-Luciferin and the amphipathic small molecule ¹⁸F-FLT. Different-sized MR and PET contrast agents demonstrated the gradual increase in leakiness of the blood tumor barrier during metastatic progression, which could guide clinicians in choosing tailored treatment strategies.

1. Introduction

Brain metastasis is the most frequent neoplasm of the central nervous system (CNS), and is ten times more common than primary malignant tumors [1, 2]. The majority of brain metastases originate from lung cancer, breast cancer or melanoma [3]. Among these, melanoma has the highest propensity to metastasize to the brain, occurring in over 50% of all melanoma patients with advanced disease [4, 5]. The incidence of brain metastases appears to be rising, due to improved detection by increased use of refined imaging modalities, and improved systemic treatments that prolong survival and control tumor burden in other organs [6-8].

Brain metastases are diagnosed late in the clinic, when they are sufficiently large to be detected by imaging, and magnetic resonance imaging (MRI) using passive gadolinium contrast enhancement is currently the method of choice [9]. At this time the prognosis is poor, as in a large study of melanoma patients with brain metastases, the median survival was 3.8 months [10].

Patients with brain metastases are treated with surgery, radiation therapy and chemotherapy. Surgery is performed on patients with single brain metastases and controllable systemic disease [10], in order to relieve pressure effects and achieve local tumor control [11]. Palliative, whole brain radiation therapy (WBRT) is the preferred treatment for most patients, since more than 70% of the patients have multiple metastases at the time of diagnosis [11]. Control of neurological symptoms may be achieved in 70-90% of the patients after WBRT. Stereotactic radiosurgery (SRS) is a method for delivering focused single high-dose irradiation, using a Gamma Knife or a dedicated linear accelerator. Local growth control of single brain metastases is high, and SRS is also effective on tumors that are regarded resistant to WBRT, such as brain metastases from melanoma [11]. SRS can safely be performed repeatedly on new brain metastases, due to low radiation doses to nearby brain tissue. The role of chemotherapy for brain metastases is limited [12]. Temozolomide has been the most promising drug, either used alone, in combination with other cytotoxic or biological drugs, or combined with WBRT [12]. Most of the chemotherapeutic agents used today are too large or hydrophilic to cross an intact (non-leaky) blood-brain barrier (BBB), making the brain a “sanctuary” site for metastasis [6, 13].

Medical treatment of brain metastases is to date controversial, as survival is only modestly improved [12]. The dismal prognosis after treatment strongly suggests that early detection is needed for improved therapy and hence better patient outcome.

Also, the efficacy of treatment has been shown to diminish in larger lesions as tumor cells create an altered local environment via their interactions with astrocytes. Thus, treatment strategies that are highly effective in vitro often fail once this altered microenvironment is established [14, 15]. Here, in order to enable early detection and characterize the BBB to improve treatment of small metastases, we set out to evaluate accumulation and transport kinetics of contrast agents of varied size and lipophilicity in serial studies of brain metastasis development.

In the presence of brain tumors, the BBB in the tumor area is commonly referred to as the blood-tumor barrier (BTB). There is a heterogeneous permeability (leakiness) to different-sized molecules through the BTB, with many metastases showing little or no permeability at all [16-20]. Thus, efficient drug delivery to brain metastases may be compromised by an intact BTB around small lesions [14]. Experimental studies of BTB permeability have been restricted to employing fluorescence imaging, and recently a few high field MRI studies have been performed. Here, we provide for the first time a detailed characterization of changes in BTB permeability by performing longitudinal molecular imaging studies using a wide range of contrast agents, by incorporating a novel MRI probe and also by comparisons with PET imaging.

Very few relevant experimental brain metastatic models are available to study biological mechanisms and therapeutic limitations and potentials in small brain metastases in relation to the BTB [17, 21]. Intracardial injections of human tumor cell lines followed by intravenous injections of sodium fluorescein (MW 376D) showed that lesions had an intact BTB until they reached a diameter larger than 0.25 mm [5, 22]. In a permeability study using a mouse model of breast cancer metastasis, Percy and collaborators injected Dextran (MW 3kDa) and Gd-DTPA (MW 938D), and showed by MRI that a significant number of brain metastases were non-permeable, and that permeability developed late in tumor development [17]. Others administered Texas Red Dextran and radiolabeled chemotherapies to tumor bearing mice to demonstrate partial BTB permeability varying in magnitude within and between metastases [16].

It is evident that still very little is known about changes in BTB permeability during brain metastasis formation. Moreover, there is little data on the ability of high field MRI to improve detection of smaller brain lesions, which likely are impermeable to contrast agents. We address these issues by performing multimodal imaging studies following administration of contrast agents with various molecular weights (Suppl. Fig. 2) and thus different abilities to cross the BTB, into a well-established melanoma brain metastasis model [23]. By combining these contrast agents with bioluminescent imaging, MRI and PET imaging, we show that impermeable brain metastases as small as 100 μm may be detected by high field MRI, and the tumors become permeable to small molecular weight contrast agents upon reaching a diameter of 200 μm. However, accumulation and transport kinetics of agents in these small lesions is dependent on the agent properties. This is likely to be important as improved diagnosis and treatment of brain metastases is critical and will likely require a personalized treatment approach.

2. Materials and Methods

2.1 Cell lines and cell culture

The H1 cell line was developed in our laboratory from a patient biopsy of a human melanoma brain metastasis as previously described [24] (Suppl. Fig. 1a i-iv). Written consent was obtained from the patient before tumor material was collected. The Regional Ethical Committee (REC number 013.09) and the Norwegian Directorate of Health (NSD number 9634) approved tissue collection and biobank storage of tumor biopsies and derived cell lines. The cells were authenticated in February 2013 using the AmpFℓSTR Profiler Plus PCR Amplification Kit (Applied Biosystems) and short tandem repeat (STR) profiles were matched to the parent tumor and cross-checked with cell line profiles at www.dsmz.de.

The H1 cells were transduced with two lentiviral vectors, encoding Dendra (a GFP variant) and Luciferase to obtain the H1_DL2 cell line (Suppl. Fig 1a v, and Supplementary Material). Flow cytometric isolation of cells by GFP expression was performed (BD FACSAria, Becton Dickinson, Franklin Lakes, NJ, USA) (Suppl. Fig 1A vi and vii). The transduced H1_DL2 cell line was proven positive for Luciferase activity by in vitro bioluminescence imaging (BLI), and then used in all experiments.

The cells were grown in DMEM supplemented with 10 % heat-inactivated newborn calf serum, four times the prescribed concentration of non-essential amino acids, 2 % L-Glutamine, penicillin (100 IU/ml), and streptomycin (100 μl/ml) (BioWhittaker, Verviers, Belgium). The cells were kept in a standard tissue culture incubator at 37°C (100 % humidity, 5 % CO₂). The growth medium was exchanged twice a week.

2.2 Animal model

All animal studies were conducted under a protocol approved by the University of California Davis, Institutional Animal Care and Use Committee. Eight week old female NOD/SCID mice (NOD.CB17-Prkdc^scid/NcrCrl) weighing 19-22g were purchased from Charles River Laboratories International (Wilmington, MA, USA). During all experiments, the mice were anesthetized with 3% isofluorane (in oxygen, flow 2L/min) and maintained at 1.5% isofluorane (in oxygen, flow 2L/min).

2.3 Tumor cell injections

The mice received a subcutaneous injection of 0.05 ml Buprenorphine hydrochloride (Buprenex, 0.05-0.1mg/kg; Cardinal Health, Elk Grove, CA) for prolonged pain relief post injection. 5 × 10⁵ H1_DL2 cells suspended in 0.1 mL PBS were slowly injected into the left cardiac ventricle by freehand using a 30G insulin syringe (Omnican50, B. Braun Melsungen AG, Melsungen, Germany), as described previously [25] (Suppl. Fig. 1a viii). Tumor development was then followed by multimodal imaging weekly for 6 weeks, as described below (Suppl. Figs. 1 a ix and 1b).

2.4 Bioluminescence imaging (BLI)

BLI was performed 10 min after tumor cell injections to determine possible inoculation failures, and weekly over 6 weeks to study systemic tumor development (29 mice, see Suppl. Fig. 1b). Anesthetized mice were depilated and given an intraperitoneal injection of 150 mg/kg D-luciferin Firefly (Gold Biotechnology, St. Lois, MO), 5 min prior to whole body imaging with a Xenogen Ivis 100 Small Animal Molecular Imager (Xenogen Corporation, Alameda, CA). Photons were collected for 3 s. Two healthy animals (no tumor cells injected) were also imaged using the same BLI procedure in order to show that light signals were not observed in these animals. The images were analyzed using Living Image v2.50 software (Xenogen Corporation). A circular 4.4 cm² region of interest (ROI) was placed over the head area of each mouse (both on dorsal and ventral images, see also Fig. 1c), and the total photon counts were calculated every week between weeks 1-6. The increase in total photon counts was analyzed using GraphPad Prism v6 (Graphpad Software, Inc., La Jolla, CA).

Figure 1.

Bioluminescence imaging (BLI) of metastatic cell growth in brains of NOD/SCID mice. (a) BLI of ventral (left) and dorsal (right) side of the animals, one week after intracardiac injections of tumor cells. (b) BLI of ventral (left) and dorsal (right) side of the animals, two weeks after injections. (c) BLI of ventral (left) and dorsal (right) side, four weeks after injections. The total photon counts/s were calculated within a ROI placed over the head area of each mouse (an example of ROI size and placement is shown on the ventral figure). (d) The total BLI signal was measured each week, on the ventral side (left) and on the dorsal side (right) (mean ± SEM).

2.5 Magnetic resonance imaging (MRI)

MRI was performed using a Bruker Biospec 7 Tesla (7T) small animal scanner (Bruker BioSpin MRI, Ettlingen, Germany) equipped with either a circular mouse head transmit/receive coil or a cross coil configuration with a rat body linear resonator for transmit and a four channel mouse brain phased array for receive. Images were acquired using ParaVision 5.1 (Bruker BioSpin MRI).

The development of brain metastatic tumor burden was determined by T1 weighted (T1w) MRI. The difference in tumor leakiness between Gd-HPDO3A (Prohance®, 0.5 μmol/g, corresponding to a gadolinium injection of 0.5μmol/g mouse, longitudinal relaxivity r₁ = 3.904 mM/s at 7T) and a newly synthetized Gadolinium contrast agent with 3 Gd(III) chelates, termed C3 (MW 2.066kD, 0.167 μmol/g, corresponding to a gadolinium injection of 0.5μmol/g mouse, longitudinal relaxivity r₁ = 16.2 mM/s at 7T) [26], was assessed by T1 mapping at weeks 5 and 6 (47 animals using Prohance, and 18 animals using C3, see also Suppl. Fig. 1b). Differences in tumor kinetics using either Prohance or C3 were also studied by T1w MRI (see Supplementary Material) and differences in accumulation of these tracers was examined over an extended period with a T1 mapping protocol. Healthy animals (no tumor cells injected) were also imaged using the same T1w MRI procedures in order to show that hyperintensive areas were not observed in the normal brains.

2.6 Positron emission tomography (PET) imaging

The syntheses of ¹⁸F-FLT and ⁶⁴Cu-BSA followed a previously reported method (see Supplementary Material). In vivo PET imaging was performed with a micro-PET scanner (Focus 120, Siemens Medical Solutions, Inc, Malvern, PA).

Animals that received ¹⁸F-FLT were imaged immediately after tail-vein injection and data were acquired for 30 min (10 tumor animals and 4 control animals). After the PET scans, the mice were euthanized by cervical dislocation and the brains were harvested and weighed. Radioactivity was measured using a Wallac Wizard 1470 Automatic Gamma Counter (Perkin-Elmer Life Sciences, Waltham, MA). Animals receiving ⁶⁴Cu-BSA were scanned at 0, 3, 6, 18 and 24 hours after injection and data were acquired for 30 min (12 tumor animals and 6 control animals). All PET images were analyzed using AsiPro software (Concorde Microsystems Incorporated, Knoxville, TN). ROIs were drawn in the brain on consecutive coronal PET images (total ROI volume around 0.25cm³) and the percent of the injected dose per volume (%ID/cm³) within the ROI volume was generated.

2.7 Histology and immunohistochemistry

Mouse brains were stained with haematoxylin and eosin (H&E) to evaluate morphology, HMB45 to identify melanocytic cells and CD31 to identify blood vessels in tumor and normal brain (see Supplementary Material).

2.8 Statistics

The statistical analyses were conducted using Graphpad Prism v6 for Mac OS X (Graphpad Software, Inc., La Jolla, CA). Differences in clearance of contrast agents from tumor (Fig. 4e) and PET tracer accumulation (Figs. 6e-f) were statistically tested using one-way ANOVA, followed by Bonferronis multiple comparisons test. For all other statistical testing, an unpaired two-tailed T-test with Welch’s correction assuming unequal variances was used. For all tests, p ≤ 0.05 was considered statistically significant.

Figure 4.

Evaluation of tumor kinetics by T1w MRI with Prohance or C3 contrast enhancement. (a) Image example of a mouse imaged at 5 weeks, after administering 0.5 μmol/g Prohance. Scale bar = 2mm. (b) Image example of the same mouse as in (a) imaged at 5 weeks, after administering 0.167 μmol/g C3. The time interval between two contrast injections was always at least 24 hrs. (c) An example of a measured T1 kinetic curve, after administering 0.5 μmol/g Prohance (left), and an example of a measured T1 kinetic curve, after administering 0.167 μmol/g C3 (right). (d) Apparent permeability into tumor, after administering 0.5 μmol/g Prohance (left), and apparent permeability out of tumor, after administering 0.167 μmol/g C3 (right) (mean ± SEM). (e) Clearance of Prohance from tumor, after administering 0.5 μmol/g Prohance (left) and clearance of C3 from tumor, after administering 0.167 μmol/g C3 (right) (mean ± SD, and min-max). (d-e) n.s.: not significant, *: p < 0.05, **: p < 0.01, ***: p < 0.001, ****: p < 0.0001.

Figure 6.

Evaluation of PET tracer accumulation in the mouse brain during metastatic development. (a) Representative coronal ¹⁸F-FLT PET images of a control animal, a tumor animal after 4 weeks and a tumor animal after 6 weeks. Scale bar = 5mm. (b) For each animal, regions of interest within the brain were drawn on 14 consecutive coronal images (giving a total volume of 0.24 ± 0.02 cm³), and the ¹⁸F-FLT activity (measured in %ID/cm³) was determined. *: p<0.01, **: p<0.001. (c) MR image of a tumor mouse 4 weeks after cell injections (left), and a corresponding ⁶⁴Cu-BSA PET image (right). Scale bar left = 2mm, scale bar right = 5mm. (d) MR image of the same tumor mouse as in (c), 6 weeks after cell injections (left), and a corresponding ⁶⁴Cu-BSA PET image (right). Scale bar left = 2mm, scale bar right = 5mm. (e) Measurements of ⁶⁴Cu-BSA activity in the tumor brains after 4 weeks and (f) 6 weeks after injection of tumor cells. (e-f) n.s.: not significant, **: p<0.01, ****: p < 0.0001. Example of %ID/cm³ in blood (g) and tumor (h) after 6 weeks is also shown.

3. Results

3.1 BLI shows early tumor development

BLI is regarded as a sensitive imaging technique, allowing detection of BBB integrity [27] as well as detection of smaller tumor cell numbers than by other imaging modalities [28]. Here BLI was used for the assessment of longitudinal brain metastasis growth, especially for early tumor development. BLI was first used to confirm successful injections, as seen by an evenly distributed BLI signal within the animals (Sup. Figure 3a). For all animals, BLI confirmed tumor cells localized within the brain after one week (Figure 1a), and a gradual increase in total photon counts/s was seen at week 2 (Figure 1b), week 4 (Figure 1c) and week 6 (Sup. Figure 3c). The total photon counts/s decreased during the first week after injection (Figure 1d), presumably due to the loss of cells lodged within the brain vasculature through apoptosis. After the first week there was a logarithmic increase in total photon counts/s both when measured ventrally and dorsally (Figure 1d). Animal survival averaged 7 weeks after tumor cell injection, and the mice exhibited over 10% weight loss during the experiments (Sup. Figure 3b).

3.2 Melanoma brain metastases have distinct growth patterns

To study the growth characteristics of the melanoma cells in more detail, we followed the development of metastases from the single cell level to macroscopic lesions by histology. After 1 week we detected single HMB45-positive melanoma cells in the brain parenchyma primarily located in close vicinity to small capillaries (Figure 2a, asterisk). After 2 weeks post-injection we observed melanoma cells arranged in small clusters in the brain (Figure 2b). Focally, a more diffuse vessel co-optive tumor growth was seen after 3 weeks post-injection (Figure 2c). After 4 weeks, formation of solid tumors was observed (Figure 2d). The number of blood vessels was significantly higher in normal cortex than in the medulla. Further, the microvessel density in the tumors after 4 and 6 weeks was considerably lower than that observed in normal brain tissue (Figures 2e-2g). The tumor microvessel diameter after 4 weeks was comparable to the normal brain, but was significantly increased after 6 weeks (Figure 2h). The mean vascular area fraction at 4 weeks was lower than the normal brain, but increased significantly compared to normal brain tissue after 6 weeks (Figure 2i).

Figure 2.

Histological characterization of melanoma brain metastasis. (a) HMB45 stained histological section showing a single melanoma brain metastatic cell (asterisk) in the mouse brain parenchyma, 1 week after intracardiac injections of the H1_DL2 cell line. Arrowheads indicate an endothelial cell. Scale bar = 100 μm. (b) HMB45 stained histological section showing a small cluster of melanoma brain metastatic cells in the mouse brain parenchyma after 2 weeks. Scale bar = 50 μm. (c) HMB45 stained histological section showing a melanoma brain metastasis in the mouse brain after 3 weeks. Scale bar = 50 μm. (d) H&E stained histological section of a melanoma brain metastasis in the mouse brain, 4 weeks after intracardiac injections of the H1_DL2 cell line. Scale bar = 50 μm. (e) CD31 stained section of the same tumor as in (d), showing vasculature. Scale bar = 50 μm. (f) CD31 stained histological section of a melanoma brain metastasis in the mouse brain after 6 weeks. Scale bar = 50 μm. (g) Evaluation of microvessel density in H1_DL2 brain metastases, 4 weeks (n=4 mice) and 6 weeks (n=5 mice) after intracardiac injections of tumor cells (mean ± SEM). (h) Evaluation of microvessel diameter after 4 and 6 weeks. (i) Evaluation of vascular fraction after 4 and 6 weeks. n.s.: not significant, **: p < 0.001, ***: p < 0.0001.

3.3 T1w MRI detects brain lesions down to 100 μm in diameter, and Prohance enhanced MRI visualizes tumor leakiness and increased tumor burden

Tumors could not be visualized with T1w MRI (without or with contrast) at week 3 (data not shown). However, T1w MRI was able to detect tumors after 4 weeks (Figures 3a-c), and we could follow the longitudinal growth of the individual metastases at week 5 and week 6 (Figures 3d-f). There was an exponential increase in the total tumor burden in the mouse brain from week 4 to 6, increasing from 1.0 mm³ to 22.1 mm³ during this time period (Sup. Figure 4b).

Figure 3.

Evaluation of tumor burden and tumor leakiness using MRI with Prohance contrast enhancement, and image subtraction. (a-c) Representative MR images of the brain of a mouse, 4 weeks after tumor cell injections. Examples of two non-leaky tumors are shown (arrows). T1 weighted images were obtained (a) before (pre) and (b) after (post) injections of the Prohance contrast agent. (c) Subtraction images (post-pre) were calculated in MATLAB. (d-f) MR images of the brain of the same mouse as shown in (a-c), 6 weeks after tumor cell injections. The same two tumors as shown in (a-c) have now become leaky (arrows). Increased tumor burden (in both numbers and size) and leakiness is seen. (g) Mean number of non-leaky and leaky tumors detected in the mouse brains at weeks 4, 5 and 6. (h) Mean diameter of non-leaky and leaky tumors in the mouse brains at weeks 4, 5 and 6, based on measurements of in total 1208 tumors. n.s.: not significant, **: p < 0.001, ***: p < 0.0001.

The mean number of brain metastases that were detectable by T1w MRI in each animal was 26.4 tumors at week 4, 74.0 tumors at week 5 and 89.0 tumors at week 6 (Sup. Figure 4a). Of these, the numbers of permeable tumors were 1.0 (3.8%) at week 4, 20.1 (27.2%) at week 5 and 48.1 (54.0%) at week 6 (Figure 3g, right), indicating a breakdown in the BTB for Prohance from approximately week 5 and onwards.

In total, 1208 measurements of tumor leakiness were performed on subtraction images. The number of leaky tumors was determined by counting the hyperintensive areas visible on the subtraction images (see Sup. Material). The smallest impermeable tumors (around 100 μm) were detected at week 4 (Figure 3h, left), the numbers peaked at week 5 (mean of 53.9 tumors, Figure 3g, left), and their mean diameters increased significantly from 250 μm to 452 μm (Figure 3h, left). The number of leaky tumors increased from week 4 to week 6 (Figure 3g, right), with the mean diameter increasing significantly from 577 μm at week 5 to 702 μm at week 6. The smallest leaky tumors detectable by MRI were around 200 μm in diameter (at week 5, Figure 3h, right).

3.4 Tumor kinetics as evaluated by T1w MRI show differences in the apparent permeability out of the tumors for Prohance and C3

We evaluated two contrast agents: Gd-HPDO3A (Prohance, MW 559D) and a multi-gadolinium agent C3 (MW 2.066kD), with a 4.1 times higher longitudinal relaxivity at 7T. T1 map values were calculated inside circular ROIs covering the whole brain lesions.

In order to ensure that the enhanced accumulation represented extravasation into the brain tumor interstitium, we evaluated the kinetics of accumulation (Figure 4) for the two gadolinium contrast agents, for example, the lesion imaged with Prohance and C3 in Figs. 4a and 4b, respectively. The arterial input function was similar for the two agents, where a bi-exponential fitting of the agent concentration in blood, i.e. $C(t)=A{e}^{-k}{}_{\mathit{1}}^t+B{e}^{-k}{}_{\mathit{2}}^t$, resulted in A=7.14±0.57 (mM), k₁=0.039±0.0043 (/s), B=0.23±0.15 (mM), k₂=-0.00031±0.00026 (/s) for Prohance, and A=1.95±0.073 (mM), k₁=0.045±0.0037 (/s), B=0.040±0.014(mM), k₂=-0.00082±0.00019 (/s) for C3.

Although the T1w signal intensity increased rapidly within the lesion for both C3 and Prohance, the rate of clearance differed with the intensity decreasing more slowly for C3 (Figure 4c, right), as compared with Prohance (Figure 4c, left). Apparent permeability into the smaller lesions (diameters less than 500 μm) was significantly greater for Prohance than for C3 (Figure 4d left). The data also showed a trend towards increased apparent permeability into larger tumors when using Prohance compared to C3, although these data were not statistically significant. Apparent permeability was also greater for larger lesions imaged with C3, compared to lesions smaller than 500 μm imaged with C3 (Figure 4d, left). The apparent permeability of the contrast agent out of the tumor was also significantly larger for Prohance than for C3, regardless of tumor size (Figure 4d, right).

Finally, we evaluated the clearance of each contrast agent over an extended period (Figure 4e). We found that C3 remained in the tumors longer than Prohance (164 min versus 49 min). Tumor T1 was not significantly different between 0 and 2 hr for C3 (p = 0.16), while for Prohance T1 significantly increased in that same time period (p = 1.98 × 10^-11). Moreover, for C3 the tumor T1 values at 0, 2, 4, and 6 hr post injection were significantly less than the pre injection T1 values. Alternatively for Prohance, the tumor T1 values at 0 and 2 hr post injection were significantly less than the pre injection T1 values, but the tumor T1 values at 4 and 6 hr post injection were not significantly less than pre injection values. At 0 hr, we found no significant difference in tumor T1 values between animals receiving Prohance and those receiving C3. However, at all later times (2, 4, and 6 hr), we found tumor T1 values to be significantly lower in animals receiving C3 compared to those receiving Prohance.

3.5 Gd-enhanced T1 mapping shows increased tumor leakiness from week 5 to week 6 after tumor cell injections

The longitudinal development in tumor leakage was followed by T1w imaging and T1 mapping 5 weeks (Figures 5a-d) and 6 weeks (Figures 5e-l) after tumor cell injections. At week 5, brain metastases were clearly visible on T1w MR images, both before (Figure 5a) and after (Figure 5c) Prohance enhancement. However, the T1 maps after contrast enhancement (Figure 5d) showed stronger hypointense lesions, compared to T1 maps before contrast injections (Figure 5b). At week 6, T1w MRI obtained before (Figure 5e) and after (Figure 5g) Prohance injections verified tumor growth compared to week 5. This was also confirmed by T1 maps acquired before (Figure 5f) and after (Figure 5h) contrast injections. Similar results were also observed for C3-enhanced lesions at week 6 (Figures 5i-l).

Figure 5.

Evaluation of tumor leakiness using MRI with Prohance or C3 contrast enhancement and T1 mapping. Examples of two 200 μm diameter lesions are shown (arrows). (a-d) Image examples of a mouse imaged at week 5, before and after administering 0.5 μM Prohance. (a) T1 weighted MR image before contrast injection. Scale bar = 2mm. (b) T1 map image before contrast injection. (c) T1 weighted MR image after administering 0.5 μM Prohance. (d) T1 map image after administering 0.5 μM Prohance. (e-h) The same mouse as in (a-d), imaged at week 6, before and after administering 0.5 μM Prohance. The same tumors as in (a-d) are also shown (arrows). (e) T1 weighted MR image before contrast injection. (f) T1 map image before contrast injection. (g) T1 weighted MR image after administering 0.5 μM Prohance. (h) T1 map image after administering 0.5 μM Prohance. (i-l) The same mouse as in (a-d), imaged at week 6, before and after administering 0.167 μM C3. The same tumors as in (a-d) are also shown (arrows). (i) T1 weighted MR image before contrast injection. (j) T1 map image before contrast injection. (k) T1 weighted MR image after administering 0.167 μM C3. (l) T1 map image after administering 0.167 μM C3. (m) Comparison of T1 map values between Prohance and C3 contrast agents, at weeks 5 and 6 (mean ± SEM). The tumors were divided into two groups according to diameters; smaller than 400 μm or larger than 400 μm. For Prohance 23 tumors smaller than 400 μm and 24 tumors larger than 400 μm were measured. For C3 8 tumors smaller than 400 μm and 10 tumors larger than 400 μm were measured. Pre = T1 values before contrast injections, Proh = T1 values after administering 0.5 μM Prohance, C3 = T1 values after administering 0.167 μM C3. n.s. = not significant, * = p < 0.01, ** = p < 0.001, *** = p < 0.0001.

At week 5 and within 30 min after injection, T1 map values for lesions smaller than 400 μm in diameter decreased from 2100 ms to 1800-1900 ms after injections of either Prohance or C3 contrast (Figure 5m, left). At this early time after injection, the differences in T1 map values between the two contrast agents were not statistically significant. For lesions larger than 400 μm in diameter, the T1 map values dropped from 1900 ms pre contrast to approximately 1700 ms after contrast injections. At week 6, T1 map values for tumors smaller than 400 μm in diameter decreased from 2000 ms pre contrast to 1800 ms post contrast, while the T1 values for lesions larger than 400 μm dropped from 1900 ms to 1500 ms. (Figure 5m, right). As shown in Figure 4, the differences in contrast agent kinetics are greatest at later time points as a result of the lower apparent permeability of the larger agent out of the tumor.

3.6 PET ¹⁸F-FLT tracer provides early indicator of accumulation and validates MRI

At week 4, an accumulation of ¹⁸F-FLT in the brains of tumor mice could be detected, where the accumulation across the entire brain had increased from 0.65 %ID/cm³ in control animals to 0.8 %ID/cm³ in mice with metastatic lesions. The %ID/cm³ continued to increase at week 6 (Figure 6a) to reach 1.7 %ID/cm³. The increase at both weeks 4 and 6 was statistically significant compared to the controls (Figure 6b). The ex vivo bioluminescence and biodistribution study confirmed accumulation of ¹⁸F-FLT at week 6 (Sup. Figure 5a). When the PET and MR images where overlaid on one another, accumulation of ¹⁸F-FLT within the metastatic brain tumors could be visualized (Sup. Figure 5b) at 6 weeks after injection.

3.7 Accumulation of ⁶⁴Cu-BSA occurs late in tumor development

Over the six weeks of imaging, the majority of the metastases could not be resolved with PET and therefore accumulation was estimated based on measurements across the entire brain. Accumulation of ⁶⁴Cu-BSA could not be detected by PET imaging after 4 weeks (Figures 6c, 6e). However, the tracer extravasated at week 6, as indicated by a statistically significant increase in the measured %ID/cm³ tissue at all time points (Figures 6d, 6f, 6h) although after accounting for the circulating agent the difference in accumulation was small (e.g. 0.18 vs 0.15 %ID/cm³ for tumor vs control animals at 24 hours after injection). The derived average apparent permeability of ⁶⁴Cu-BSA at the 6 week time point into and out of the tumor was 4.44×10^-8 (cm/s) and 2.97×10^-6 (cm/s), respectively (Figures 6g, 6h).

4. Discussion

The BTB plays an important role in the development and treatment of brain metastasis. Tumor cell attachment to the endothelium as well as extravasation and subsequent tumor cell growth is regulated by the brain vasculature [29, 30]. The delivery of chemotherapy is ineffective in the treatment of patient brain metastases, due to an often intact vascular barrier in the brain with low passive transcellular permeability [16, 19, 31, 32]. Also, the endothelial cells of the brain vasculature express active efflux drug transporters, which limit the uptake of most anticancer drugs [33, 34]. An improved ability to detect small tumors and more detailed information on the changes in permeability of the BTB over time is needed to improve clinical management of cancer patients, especially in early brain metastatic development. By using multimodal imaging, we show that molecules with a wide range of sizes (244D-65.55 kDa) are able to leak through the BTB at various stages of tumor development. We demonstrate that larger molecules (such as albumin) cannot accumulate until the final stages of tumor progression (week 6), when the mean tumor diameter of leaky tumors is approximately 700 μm (Figure 3h, right) To our knowledge, this is the first experimental study utilizing a combination of BLI, PET and MRI to determine and quantify alterations in BTB permeability over time.

The detection limit for small, non-leaky brain metastatic lesions can be improved by MRI

BLI combined with histology provided a highly sensitive gold standard for animal imaging studies with MRI and PET. BLI confirmed the presence of viable tumor cells in the brain after 1 week. The signal intensity was lowest at that time point, suggesting that a subset of metastatic tumor cells were able to survive, attach to the endothelial cells and extravasate into brain parenchyma. The histology confirmed single tumor cells localized within the perivascular area at week 1, which suggest that the microvasculature was at least partly permeable to D-Luciferin at that time. Our results are in line with other studies showing extravasation of the vast majority of the tumor cells 3-10 days after tumor cell injections [6, 35, 36]. Our findings show, however, that we were not able to detect the lesions with MRI until week 4 of the study.

Our work shows that T1w 7T MRI is able to detect non-leaky brain metastases down to approximately 100 μm in diameter (~40-50 tumor cells, as estimated by histology). As we did not perform a whole-brain histological study of the number of small metastases present, we could not estimate the percentage of non-permeable brain lesions that were detectable by MRI. The presence of melanin molecules within the melanoma cells, as seen in our model, may contribute to the observed paramagnetic effect on signal intensity [37, 38]. However fatty brain tumors, and brain lesions which are hemorrhagic, protein-containing or calcified have also been reported to shorten the T1 relaxation in the clinic without the use of contrast agents [37], suggesting that our results may also be valid for high-field T1w MRI of small brain metastases from other primary malignancies.

Previously, experimental tumors of 100 μm size have been detected by using targeted MRI contrast agents. Antibodies to endothelial vascular cell adhesion molecule-1 (VCAM-1) were conjugated to microparticles of iron oxide (MPIO), and VCAM-MPIO were injected into tumor bearing mice prior to T2* weighted MRI. Brain metastases were detected 5 days after tumor cell injections (<1000 cells) [9]. A disadvantage, however, is that almost all currently available MPIOs are not approved for clinical use. Thus, alternative MRI protocols such as high-field T1w MRI without T1 contrast, could be of particular interest for early tumor detection.

The current clinical detection limit for brain metastases using MRI is approximately 2-5 mm, when the tumors become permeable and the tumor volume approaches the MR resolution [39]. Recently, commercial high field (7T) human MR scanners have become available and the increased field strength offers a higher signal to noise ratio that allows for smaller voxel dimensions when imaging brain structures [40]. Although the exact MR imaging parameters used in this study are not feasible for human whole brain MRI, 7T human scanners may enable screening for small metastases in the near future. Indeed, 7T scanners have already been used for scanning of astrocytomas as well as brain metastasis [41, 42]. Simulations with a human Philips 7T MR have shown that a T1w whole brain image with a voxel size of 0.24 × 0.24 × 3 mm with a 24cm × 24cm FOV would require approximately 33 min and does not exceed specific absorption rate limits. This scan time could easily be halved using partial Fourier, SENSE, or GRAPPA [43-45]. Following a whole brain scan, a reduced field of view scan of suspect lesions could enable MRI kinetic analysis, similar to that performed in this paper to verify altered permeability within these lesions [46].

The resolution in our MR studies was 78×78×1000 μm, while the PET resolution was 1.5×1.5×1.5mm (based on aposteri reconstruction). Thus, a 200 μm lesion (~350-400 tumor cells) at the time of detection filled 2/3 of the MRI voxel but 1/800 of the PET voxel. For the MRI contrast imaging, we injected 10 μmoles of gadolinium contrast. We have previously observed an accumulation on the order of 5% ID/g in larger developing lesions [47]. Assuming a peak concentration of 5% ID/cm³ within the tumor and correcting for the partial volume effect of the lesion, we would observe ~100 nmoles/cc of gadolinium with a detection limit of ~10 nmoles/cm³ of gadolinium. Therefore, 200 μm lesions accumulating gadolinium should be within our limits of detection for MRI.

For the microPET focus scanner used here, we have found that ~3.33 Bq (9 × 10^-11 Ci) is required within the resolution volume for detection in a 30-minute scan (data not shown). For our injection of 200 μCi of activity and a typical lesion accumulation of 5% ID/cm³, ~4.2 × 10^-11 Ci would accumulate within a 200 μm lesion. Therefore, for a PET scanner with this relatively high resolution assuming one lesion is present within the resolution volume, detection of ~260 μm lesions is feasible.

Pharmacokinetic modeling shows extensive changes in tumor permeability for various sized contrast agents during tumor progression, which has therapeutic implications

Very little is known about variations in permeability during brain metastasis progression. However, it is likely that the change in tumor permeability is a dynamic process in itself [17], necessitating longitudinal monitoring to study the permeability changes. Effective therapeutics must both penetrate the BBB and be retained for a sufficient time interval to produce a therapeutic response. Candidate therapeutics include small molecules, proteins, antibodies and fragments and nanoparticles ranging from effective diameters of nanometers to hundreds of nanometers. A few groups have previously developed animal models of brain metastasis and performed studies of tumor development and BTB integrity using gadolinium-enhanced MRI [17, 48, 49], fluorescence-based imaging [5, 16, 20, 22, 50], as well as i.v. injections of Evans Blue (which binds to plasma albumin) followed by immunohistochemistry [51].In the current work, we expand on current knowledge by providing a detailed analysis of early tumor detection and disruption of the BTB in brain metastatic progression in an established and reproducible animal model of melanoma brain metastasis [23].

We found that ¹⁸F-FLT, (MW 244D, and an amphipathic log P of -0.41) could be detected in the brain as early as 4 weeks after the injection of tumor cells, corresponding to a mean tumor diameter on the order of 250 μm (Figure 3h left). Such lesions could not be resolved within the PET image; however, due to the low background of radiation within the brain the increased activity was still detected by ¹⁸F-FLT PET.

Further, we show by T1w MRI and T1 mapping after Prohance contrast injection that brain metastases down to around 200 μm in diameter start becoming permeable. This is likely due to increased leakage of already existing brain blood vessels incorporated into the growing tumors, as the histology showed decreased microvessel density (and thus no angiogenesis) within the tumors at weeks 4 and 6, but increased microvessel diameter and increased vascular fraction.

We compared Prohance to C3, a larger construct carrying three gadolinium molecules, and injecting a similar number of gadolinium molecules. Contrast agents of similar molecular weights have previously been shown to clear rapidly from the blood circulation [52-54]. The peak reduction in T1 was similar for the two treatments; however, the kinetics of clearance was significantly different. Whereas the accumulation of Prohance was similar to C3, the retention time of C3 was greater. Thus, the apparent permeability into the lesions was higher for Prohance, and the apparent permeability out of the tumor was two times larger for Prohance than C3, regardless of tumor size (2.27×10^-5 cm/s versus 1.10×10^-5 cm/s).

Our ⁶⁴Cu-BSA PET study indicates that larger molecules such as albumin are able to extravasate into tumor tissue late in metastasis development with an effective mean apparent permeability out of the lesion of 2.98×10^-6 cm/s. We also evaluated the extravasation of gadolinium-labeled liposomes in this model (data not shown), however, accumulation was not detected in the small metastases studied here even at the final time point. Taken together, the results obtained in this study indicate that adjuvant treatment with small molecular weight therapeutics (such as temozolomide or similar) may be undertaken for lesions with diameters as small as 100 μm, yet, methods to open the BBB in these small lesions may ultimately be required to improve treatment efficacy particularly with biological therapeutics.

It was beyond the scope of our work to perform a comparative biodistribution study of all of the different tracers applied here; given the relatively small volume of lesions the systemic biodistribution and pharmacokinetics were not expected to differ from those in previous studies. Interestingly, in these studies of small metastases, all tracers, including albumin, cleared from the lesions (and brain) by 24 hours after administration. This is significantly different than observed for ⁶⁴Cu-BSA in larger malignant tumors where much of the BSA tracer remains present for more than 48 hours [47, 55, 56]. However, the rapid clearance of gadolinium-based contrast agents we observed in all normal organs including the liver, corresponds to results reported by others [52, 53, 57, 58].

While gadolinium enhancement aids the delineation of a brain metastasis, the complexity of the BBB implies that charge, lipophilicity, binding to plasma proteins, and size of a molecule all must be taken into account when determining permeability of a drug through the BBB. It is known that the enhancement of tumor with gadolinium may not fully reflect BBB permeability, as the intact BBB prevents passage of ionized, water-soluble molecules larger than 150 D, and it has a net negative charge on the luminal surface, which repels negatively charged ionic compounds [59]. Gadolinium agents are in general non-ionic, while many of the chemotherapeutic agents are charged and have larger molecular weights, and the drugs may easily bind to albumin and other large plasma proteins, thus preventing them from crossing the BBB. Our study showed that albumin extravasated very late in tumor development, likely due to increased disruption of the BTB. Further, the lipophilicity of a drug influences its delivery over the BTB, as an increase in lipophilicity enhances the ability to cross the BTB more efficiently [60] and this was also apparent here as the amphipathic ¹⁸F-FLT accumulated early in metastatic development.

Clinical significance of our findings

Common treatments of brain metastases include surgery, whole brain radiation therapy (WBRT) and stereotactic radiosurgery (SRS) [61], as well as chemotherapy [62]. Early detection and quantification of the permeability characteristics of micro-metastases can inform decisions on the choice of treatment. The use of SRS alone for patients with a limited number of metastases has increased, and has shown local tumor control and patient survival comparable to SRS + WBRT or WRBT alone. Also, the use of WBRT often leads to neurological side effects [63, 64].

There is now clinical evidence suggesting that temozolomide (TMZ, MW 194D) may be an effective adjuvant treatment of microscopic brain metastases when combined with SRS [63]. Early detection of small metastases as shown in our study, would then be of vital importance when deciding which patients should receive such combined therapy. As TMZ crosses the intact BBB, the drug can be used early in patient treatment. Large MW drugs such as ipilimumab (MW 148 kD) have shown to improve patient survival when combined with SRS [65]. Our study clearly shows that ⁶⁴Cu-BSA leaks into tumor parenchyma late in tumor development, suggesting that large MW drugs should preferably be used on larger and well-established brain metastases.

Typical accuracy of SRS is on the order of 1 mm [66], providing a means to safely deliver a focused, single high-irradiation dose to small metastasis. Other techniques, such as focal opening the BBB with contrast-enhanced ultrasound can concentrate chemotherapeutics with similar precision and may provide another method to treat these tumors after early detection [67].

In conclusion, earlier detection of brain metastases will likely change the therapeutic strategies of patients, both for current treatments as well as new and more targeted therapies. We have shown that T1w high field MRI without contrast can be used to detect 100 μm-sized metastases. Further, detailed knowledge on when the BTB is disrupted, along with knowledge of the characteristics of a drug (size, charge, lipophilicity, binding to plasma proteins) will aid clinicians in deciding which chemotherapies should be used at various stages of tumor progression. We have shown that the BTB is relatively leaky for small contrast agents (MW below 300 D) at early stages and permeable to MR contrast agents (MW 0.566kD and 2.066 kD) at later stages of brain metastatic development.