Intrinsic Susceptibility MRI Identifies Tumors with ALKF1174L Mutation in Genetically-Engineered Murine Models of High-Risk Neuroblastoma

The early identification of children presenting ALKF1174L-mutated neuroblastoma, which are associated with resistance to the promising ALK inhibitor crizotinib and a marked poorer prognosis, has become a clinical priority. In comparing the radiology of the novel Th-ALKF1174L/Th-MYCN and the well-established Th-MYCN genetically-engineered murine models of neuroblastoma using MRI, we have identified a marked ALKF1174L-driven vascular phenotype. We demonstrate that quantitation of the transverse relaxation rate R2* (s−1) using intrinsic susceptibility-MRI under baseline conditions and during hyperoxia, can robustly discriminate this differential vascular phenotype, and identify MYCN-driven tumors harboring the ALKF1174L mutation with high specificity and selectivity. Intrinsic susceptibility-MRI could thus potentially provide a non-invasive and clinically-exploitable method to help identifying children with MYCN-driven neuroblastoma harboring the ALKF1174L mutation at the time of diagnosis.


Introduction
Neuroblastoma arises in the sympathetic nervous system during embryogenesis and is the most common extracranial solid tumor in children [1]. For the established subset of patients presenting with high-risk neuroblastoma, the current portfolio of therapeutic options has limited success, with five-year survival rates rarely exceeding 40%.
The poor clinical outcome and aggressive tumor phenotype of high-risk neuroblastoma strongly correlates with amplification of the proto-oncogene MYCN and enhanced tumor angiogenesis [2]. Recently, mutations in the anaplastic lymphoma kinase (ALK) tyrosine kinase gene have been identified in ,8-10% of primary neuroblastoma, leading to constitutive activation of the ALK protein. The most common and potent ALK mutation, ALK F1174L , is associated preferentially with MYCN amplification, a markedly poorer prognosis, and confers resistance to the promising ALK inhibitor crizotinib [3][4][5].
With crizotinib in pediatric phase I clinical trials, and other ALK inhibitor studies in development, a current challenge is to rapidly identify upfront children with high-risk ALK mutated or amplified neuroblastoma who may benefit from or become resistant to ALKtargeted therapy.
As most pediatric cancers, including neuroblastoma, originate from only a few genetic anomalies during development, they are amenable to genetically engineered mouse (GEM) modeling approaches. GEM models of neuroblastoma, such as the Th-MYCN murine model, which develop spontaneous tumors mirroring the major pathophysiological, genetic and radiological features of high-risk MYCN-amplified childhood neuroblastoma, represent clinically-relevant tools for the study of neuroblastoma biology and response to novel therapeutics [6][7][8]. The development of GEM models co-expressing ALK F1174L and MYCN to the neural crest, such as the Th-ALK F1174L /Th-MYCN model, have recently been used to demonstrate how the ALK F1174L mutation potentiates the oncogenic activity of MYCN, and that ALK F1174L acquired resistance to crizotinib can be overcome through inhibition of key cellular modulators of n-myc [9][10][11].
In this study, we hypothesized that the presence of the ALK F1174L mutation results in a phenotypic difference in hemo-dynamic vasculature in tumors in the Th-ALK F1174L /Th-MYCN model, and which can be evaluated using intrinsic susceptibility magnetic resonance imaging (MRI). The aim of our study was to demonstrate that quantitation of the transverse relaxation rate, R 2 *, and changes in R 2 * induced by breathing 100% oxygen, DR 2 * oxygen-air , could discriminate between tumors arising in Th-ALK F1174L /Th-MYCN and Th-MYCN mice, and that intrinsic susceptibility MRI could thus potentially provide a non-invasive and clinically-translatable method to help identify children, presenting MYCN-amplified neuroblastoma harboring the ALK F1174L mutation.

Ethics statement
All procedures involving animals were approved by the Institute of Cancer Research Animal Ethics Committee and the UK Home Office and carried out according to the United Kingdom National Cancer Research Institute guidelines for the welfare of animals in cancer research [12].

Animal models
The generation of the Th-ALK F1174L /Th-MYCN mice has been recently described [11]. Th-ALK F1174L /Th-MYCN and Th-MYCN mice were identified by analyzing DNA from mice tails using realtime quantitative reverse transcription polymerase chain reaction (qRT-PCR, Transnetyx Inc., Cordova, Tennessee), and mice with tumors were initially identified by palpation.

Magnetic Resonance Imaging
All the 1 H MRI studies were performed on a 7T Bruker horizontal bore micro-imaging system (Bruker Instruments, Ettlingen, Germany) using a 3 cm birdcage coil. Anesthesia was induced by an intraperitoneal 0.1 ml injection of a combination of fentanyl citrate (0.315 mg/ml) plus fluanisone (10 mg/ml) (Hypnorm, Janssen Pharmaceutical, Oxford, UK) and midazolam (5 mg/ml) (Roche, Welwyn Garden City, UK) and water (1:1:2). A nose piece was positioned for oxygen delivery.
For all the mice, anatomical T 2 -weighted coronal and transverse images were acquired from twenty contiguous 1 mmthick slices through the mouse abdomen, using a rapid acquisition with refocused echoes (RARE) sequence with 4 averages of 128 phase encoding steps over a 363 cm field of view, two echo times (TE) of 36 and 132 ms, a repetition time (TR) of 4.5 s and a RARE factor of 8. These images were used to determine tumor volumes, and for planning the intrinsic-susceptibility MRI measurements, which included optimization of the local field homogeneity. The baseline transverse relaxation rate R 2 *, sensitive to the concentration of paramagnetic species, principally deoxyhemoglobin, was quantified in tumors from Th-ALK F1174L / Th-MYCN (n = 23) and Th-MYCN (n = 21) mice, using a multiple gradient echo (MGE) sequence. MGE images were acquired from three 1 mm thick transverse slices through each tumor, using 8 averages of 128 phase encoding steps over a 363 cm field of view, and an acquisition time of 3 min 20 s. Images were acquired using 8 echoes spaced 3 ms apart, an initial echo time of 6 ms, a flip angle a = 45u and a repetition time of 200 ms. Subsequently, 100% oxygen (BOC Ltd, Guildford, UK) was delivered at a rate of 2 L/min to Th-ALK F1174L /Th-MYCN (n = 10) and Th-MYCN (n = 12) mice. After a 5 minutes equilibrium period, identical MGE images were acquired whilst the mouse continued to inhale oxygen.
All the MGE data were fitted voxelwise using in-house software (ImageView, working under IDL, ITT, Boulder, Colorado, USA) with a robust Bayesian approach that provided estimates of R 2 * and DR 2 * oxygen-air ( = R 2 * oxygen 2R 2 * air ) [13].

Histological Assessment
Formalin fixed paraffin embedded sections from Th-ALK F1174L / Th-MYCN and Th-MYCN mice were stained with haematoxylin and eosin, and visualized under light microscopy. The extent of functional vasculature in tumors from both models was assessed using the perfusion marker Hoechst 33342, and quantified as fluorescent area fractions (%), as previously described [8].

Statistical analysis
Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software Inc., La Jolla, USA). The mean values for tumor volume, mean of median values for R 2 * and DR 2 * oxygen-air , and the mean fluorescent area fractions were used for statistical analysis. R 2 * and DR 2 * oxygen-air were assumed to be normally distributed, which was confirmed using the D'Agostino-Pearson omnibus K 2 normality test [14]. Any significant difference in tumor volume, R 2 *, DR 2 * oxygen-air and the fluorescent area fractions between Th-ALK F1174L /Th-MYCN and Th-MYCN mice were identified using Student's 2-tailed unpaired t-test, with a 5% level of significance.

ALK F1174L mutation induces a differential radiological presentation of neuroblastoma
Anatomical T 2 -weighted RARE MR images revealed solid masses within the retroperitoneum in peri-renal and para-spinal abdominal regions of both the Th-ALK F1174L /Th-MYCN and Th-MYCN mice, typical of the clinical distribution and radiological presentation of human neuroblastoma ( Figure 1A) [15,16]. All the Th-ALK F1174L /Th-MYCN (n = 23) and Th-MYCN (n = 21) mice examined presented with abdominal tumors with wide range of volumes (525-3400 mm 3 for tumors in Th-ALK F1174L /Th-MYCN mice and 335-2450 mm 3 for tumors in Th-MYCN mice). Tumors in the Th-MYCN mice appeared heterogeneous, with 94% of tumors presenting with areas of hypointense signal, consistent with previous observations [8]. Tumors in the Th-ALK F1174L /Th-MYCN mice were generally hyperintense and significantly more homogeneous, with only 48% of the tumors having hypointense regions (p = 0.006, x 2 -test). On T 2 *-weighted images ( Figure 1B and C, and Figure 2A), tumors in the Th-MYCN mice were generally hypointense at longer echo times, whereas the signal from tumors in the Th-ALK F1174L /Th-MYCN remained hyperintense, providing a stark positive contrast with the surrounding organs for 74% of the Th-ALK F1174L /Th-MYCN mice (compared to 14% for the Th-MYCN mice, p,0.001, x 2 -test). ALK F1174L mutation results in a slower tumor transverse relaxation rate, R 2 * , in neuroblastoma Parametric maps revealed a heterogeneous distribution of R 2 * in tumors in both models ( Figure 2B). Quantitation of the transverse relaxation R 2 * revealed significantly slower rates in tumors in the Th-ALK F1174L /Th-MYCN cohort, compared to the faster R 2 * rate determined in tumors in the Th-MYCN mice ( Figure 2D). Baseline R 2 * detected tumors harboring the ALK F1174L mutation with a sensitivity of 90% (95% CI: 66.9-98.2) and a specificity of 81% (95% CI: 57.4-93.7). ALK F1174L mutation is associated with a stark differential BOLD MRI response to hyperoxic challenge in neuroblastoma Continuous inhalation of 100% oxygen resulted in tumors in the Th-MYCN mice demonstrating a very strong, heterogeneous increase in T 2 *-weighted image signal intensity (blood oxygen level dependent (BOLD) effect) ( Figure 2A). In contrast, tumors from Th-ALK F1174L /Th-MYCN mice showed a negligible BOLD effect on T 2 * weighted image intensity with hyperoxia. As with baseline R 2 *, parametric maps revealed a heterogeneous distribution of DR 2 * oxygen-air in tumors in both models ( Figure 2B and C).
Tumor regions showing a relatively fast baseline R 2 * remained so during hyperoxia, whereas regions of relatively slower baseline R 2 * typically showed a more marked reduction in R 2 * with 100% O 2 . The significant difference in R 2 * between the two models was lost with 100% oxygen challenge ( Figure 2E). As a consequence, tumors in Th-ALK F1174L /Th-MYCN mice demonstrated a significantly lower absolute value of DR 2 * oxygen-air than the tumors in Th-MYCN mice ( Figure 2F). Hyperoxia-induced DR 2 * oxygen-air detected tumors harboring the ALK F1174L mutation with a sensitivity of 90% (95% CI: 54.1-99.5) and a specificity of 94.1% (95% CI: 69.2-99.7).  Corresponding parametric tumor transverse relaxation rate R 2 * maps calculated during initial air breathing and after 3 minutes of continuous inhalation of 100% oxygen. C) Resulting parametric tumor DR 2 * oxygen-air (R 2 * oxygen 2R 2 * air ) maps. D) Tumor R 2 * during initial air breathing, E) tumor R 2 * after 5 minutes of breathing 100% oxygen, and F) tumor DR 2 * oxygen-air (R 2 * oxygen 2R 2 * air ) were determined from Th-ALK F1174L /Th-MYCN and Th-MYCN mice with abdominal neuroblastoma. Individual data points represent the mean of the median values determined from all three imaging slices for each animal, as well as the mean 61 s.e.m, p, Student's 2-tailed unpaired t-test with a 5% level of significance. doi:10.1371/journal.pone.0092886.g002 No correlation between tumor R 2 * and DR 2 * oxygen-air , or between either R 2 * or DR 2 * oxygen-air and tumor volume, was determined across the cohorts, indicating that the differences in intrinsic susceptibility MRI between tumors in the Th-ALK F1174L / Th-MYCN and Th-MYCN mice were independent of tumor size.

ALK F1174L mutation is associated with reduced functional vasculature in neuroblastoma
Gross examination of the tumors in situ, prior to excision, revealed an intense dark red coloration across the whole tumor in the Th-MYCN mice, in contrast to the pale appearance of tumors in the Th-ALK F1174L /Th-MYCN GEM model ( Figure 3A). Histological examination with H&E staining revealed the presence of large hemorrhagic regions filled with stacked erythrocytes in 100% of the tumors from Th-MYCN mice (n = 10), but only in 10% the Th-ALK F1174L /Th-MYCN model (n = 10, p,0.0001, x 2 -test) ( Figure 3B and C). Fluorescence microscopy of Hoechst 33342 uptake revealed homogeneously and well-vascularized tumors in both models ( Figure 3D). However, quantitation of the fluorescent area fractions revealed that tumors in the Th-MYCN mice had significantly higher uptake, consistent with the presence of more perfused, functional vasculature ( Figure 3E).

Discussion
In the current study we demonstrate that baseline R 2 * and hyperoxia-induced DR 2 * can discriminate a differential hemodynamic tumor vascular phenotype between tumors arising in Th-ALK F1174L /Th-MYCN and Th-MYCN models of high-risk, MYCN over-expressing neuroblastoma. With a sensitivity of 90% and a specificity of 81% for baseline R 2 *, and a sensitivity of 90% and a specificity of 94% for 100% oxygen-induced DR 2 *, intrinsic susceptibility MRI provides a robust method to discriminate and identify Th-MYCN transgenic mice harboring the ALK F1174L mutation.
The rapid and completely noninvasive quantitation of tumor R 2 * is suitable for the scanning of young children. Furthermore, such pediatric imaging sessions are usually performed under general anesthesia, providing an opportunity to transiently perturb the oxygen content inhaled by the patient, and enabling quantitation of tumor R 2 * under more hyperoxic conditions (from 21 to 30-40% O 2 ). Upon successful translation, intrinsic susceptibility MRI could provide a rapid method to identify children with ALK-driven tumors, enabling the stratification of children with this ultra high-risk neuroblastoma at the time of diagnosis.
The baseline transverse relaxation rate R 2 * is sensitive to the concentration of paramagnetic deoxyhemoglobin in the vascular compartment in tissue (BOLD effect). Compared to most normal tissues, tumors exhibit relatively fast native R 2 * values, a consequence of the high concentration of deoxygenated erythrocytes within the vascular compartment associated with immature, irregular and unstable microcirculation [17]. The fast baseline tumor R 2 * measured in the Th-MYCN model is consistent with the aggregation of deoxygenated erythrocytes, described as blood lakes [18], and which are characteristic of childhood neuroblastoma [19]. The significantly slower baseline R 2 * and differential intrinsic susceptibility MRI presentation of tumors in the Th-ALK F1174L / Th-MYCN mice is consistent with the absence of such blood lakes.
Inhalation of high oxygen content gases results in the rapid reoxygenation of hemoglobin, with paramagnetic deoxyhemoglobin within perfused tumor vessels being replaced by diamagnetic oxyhemoglobin, and a reduction in R 2 * [20]. Hyperoxia resulted in an overall significant reduction in R 2 * of tumors in the Th-MYCN mice, whereas the response in tumors in the Th-ALK F1174L / Th-MYCN mice was negligible, suggesting a clear difference in hemodynamic functional vasculature between the two GEM models. This was corroborated by the significant overall difference in Hoechst 33342 uptake. Interestingly, a spatially different hyperoxia DR 2 * response was apparent in tumors in the Th-MYCN mice. Tumor regions exhibiting relatively fast baseline R 2 * showed a less pronounced response with 100% O 2 breathing, also consistent with the presence of blood lakes which are typically disconnected from the perfused vascular network [21]. In contrast, tumor regions with relatively slower baseline R 2 * showed a marked response to hyperoxia, indicative of functional vasculature.
Intriguingly, tumors from the Th-ALK F1174L /Th-MYCN mice showed no clear DR 2 * response to hyperoxia, despite Hoechst Identifying ALK-Mutated Neuroblastoma with MRI PLOS ONE | www.plosone.org 33342 uptake indicating the presence of perfused blood vessels in both GEM models. This suggests a difference in vascular architecture that precludes erythrocyte delivery, but not plasma perfusion, in the tumor vessels in the Th-ALK F1174L /Th-MYCN mice, or in oxygen consumption by the tumor cells [17,22]. The extensive hemorrhage and blood lakes present in the Th-MYCN model are indicative of vessel wall instability due to rapid endothelial cell proliferation and defective pericyte coverage [23]. This implies that the tumor vasculature in the Th-ALK F1174L /Th-MYCN mice may be more mature.
Given its relationship to blood oxygen saturation and partial pressure of oxygen in and around blood vessels, quantitation of baseline tumor R 2 * is also being investigated as an imaging biomarker of hypoxia [24]. For example, slow baseline R 2 * has been shown to correlate with increased hypoxia, determined by pimonidazole staining, in chemically-induced rat mammary tumors [25]. Conversely, fast baseline R 2 * is associated with hypoxia in prostate cancer [26]. In this context, we have recently shown that neuroblastoma from Th-MYCN mice exhibiting fast baseline R 2 *, induced by a hemorrhagic phenotype, are relatively oxic, as revealed by negligible pimonidazole staining [8]. R 2 * is first of all a marker of impaired hemodynamic function and as such shares a causal relationship to hypoxia. However this relationship is ambivalent as several differential hemodynamic phenotypes can in fact lead to hypoxia, while having an opposite effect on R 2 * values. In this study the suggested tumor vascular phenotype in the Th-ALK F1174L /Th-MYCN model, which hinders the delivery of erythrocytes and causes a significantly slower R2*, may result in increased tumour hypoxia in the Th-ALK F1174L /Th-MYCN model compared with the Th-MYCN model.
Recent reports have implicated a role of the ALK F1174L mutation in tumor angiogenesis in neuroblastoma. Selective targeted inhibition of ALK resulted in a significant reduction in vascular density in xenografts derived from MYCN non-amplified SH-SY5Y cells, which harbor the ALK F1174L mutation, accompanied with a decrease in vascular endothelial growth factor (VEGF) and matrix metalloproteinases (MMPs) [27]. The expression of MMP-9 by stromal cells has previously been shown to regulate the vascular architecture in a murine orthotopic MYCN-amplified neuroblastoma xenograft model by promoting pericyte recruitment [28]. Collectively, these studies suggest that ALK and the ALK F1174L mutation contribute to tumor vasculogenesis and pericyte recruitment via the regulation of VEGF and MMP-9, leading to a dense vascular network with smaller, more stable vessels. The differential intrinsic susceptibility MRI vascular phenotype observed in tumors in the Th-MYCN and Th-ALK F1174L /Th-MYCN models reported herein demonstrates an important role for ALK F1174L in angiogenesis in MYCN-overexpressing neuroblastoma in vivo, strongly suggesting a role in vasculogenesis.
This study has some limitations. The Th-ALK F1174L mutation leads to the constitutive activation of the ALK protein in neuroblastoma [29][30][31][32], and the vascular phenotype associated with amplified or wild-type (WT) ALK expression is currently unknown. Amplification of ALK has also been shown to lead to activation of the ALK protein with, however, a 17-fold reduced kinase activity compared to ALK F1174L mutants [29]. In the absence of a Th-ALK wt GEM model, it is difficult to conclude if intrinsic susceptibility MRI would be solely able to identify ALK F1174L -mutated MYCN amplified neuroblastoma, or more generally ALK-driven MYCN amplified neuroblastoma.
There is a clear need to more deeply interrogate the role of the ALK F1174L mutation on vascular morphogenesis and architecture, which may be a major determinant of impaired drug delivery and a contributing factor to the poor prognosis of children with ALK F1174L -mutated MYCN-amplified neuroblastoma. The altered vascular phenotype may also impact on the response to antivascular therapies, including cediranib, currently being considered in clinical trials for the treatment of high-risk MYCN-amplified neuroblastoma [33]. Retrospective analysis of historical pathological samples for ALK mutations should provide sufficient statistical power to shed a light on the vascular phenotype of neuroblastoma associated with ALK amplification, and each of the rare ALK mutations, including the lethal ALK F1174L mutation.
Coupled with our recent identification of quantitation of R 2 * as a biomarker of treatment response to cediranib in the Th-MYCN GEM model [8], the present study reinforces R 2 * as a biomarker of vasculogenesis and its response to therapy in these clinically relevant GEM models. Furthermore, it provides a strong rationale for the evaluation of intrinsic susceptibility MRI for assessing any anti-angiogenic effects resulting from successful targeted inhibition of ALK signalling in ALK F1174L mutated neuroblastoma, which could ultimately used for the assessment of second generation ALK inhibitors [34,35]. To conclude, this study has provided a strong rationale for the immediate incorporation of intrinsic susceptibility MRI into forthcoming imaging-embedded clinical trials of next generation ALK inhibitors in ALK-mutated andamplified neuroblastoma.