1. IntroductionBrain metastasis (BM) is a common complication in patients with malignant tumors, especially non-small-cell lung cancer, melanoma, and breast cancer [1]. It is estimated that 20–30% of cancer patients may develop BM; however, the exact epidemiological data are not reported yet because of the possible under-estimated incidence due to (1) lack of mandatory reporting of BM status in all patients; and (2) lack of routine imaging surveillance of brain during clinical management [2]. This figure may likely increase with time, due to prolonged patient survival and increased sensitivity in detecting BM thanks to the technical development of imaging modalities.BM represents an epidemiologically distinct metastasis, compared with liver and lung metastasis, for the same primary tumor [3,4,5]. Due to the unique microenvironment of the brain, i.e., dense cellularity, blood–brain barrier (BBB), and cerebrospinal fluid (CSF), therapeutics that are effective to control primary disease are usually futile to eliminate cancer cells disseminated into the brain. This can be attributed to two major factors: (1) cancer cell evolution and/or selection (harboring novel mutations or metabolic profiling during metastasis [6]; and (2) BBB-limited drug penetration and/or tumor retention [7]. Currently available therapeutics for BM include ALK, EGFR-targeting therapies for lung cancer, for breast cancer with HER2 and for melanoma with BRAF gene, but they consist of only 18% BM cases in total [8,9]. For other BM patients without known druggable targets, immune checkpoint inhibitors showed encouraging cancer control in melanoma, triple-negative breast cancer, and NSCLC-derived BM [10]. Besides focusing on the already-existing driving mutation in primary lesions, additional mutations having developed during metastasis may be another potential target, e.g., FAM129C and ADAMTSs in BM lesions but absent in primary lung cancer [11].The BBB is constructed of a continuous layer of capillary endothelial cells connected by inter-cellular tight junctions and adherent junctions, a basement membrane, pericytes, and end-foot processes of perivascular astrocytes, whereas in the site of BM, the BBB is compromised with heterogeneous drug permeability [12].Besides strategies focusing on cancer cells, tumor stromal cells may serve as an alternative target, which are less diverse than mutative cancer cells and more homogenous among different cancer types. Uncontrollable growth, invasive capacity, and metastatic invasion into distant organs or tissues are major hallmarks of malignancies, all of which highly depend on the proliferating vascular networks [13]. Thus, targeting tumoral blood vessels has a great implication for controlling cancers by either inhibition of neoangiogenesis or disruption of existing blood vessels, resulting in tumor starvation and consequently necrosis. The former type represented by Bevacizumab (Avastin®) targets angiogenesis and have shown survival benefit to glioblastoma multiforme in a recurrent setting [14], but not in a first-line setting [15]. Thus, the latter group of anti-vascular agents includes vascular-disrupting agents (VDAs) as typically exemplified by combretastatin A-4 phosphate (CA4P), which have shown preliminary therapeutic effects in liver cancer and lung cancer [16,17,18]. However, the limited efficacy of VDAs on an orthotopic glioma model has been shown [19]. However, it would be interesting to question to what extent the existence of the BBB may impact the therapeutic effect of a VDA on BM as compared to extracranial tumors. Furthermore, exploring the efficacy of VDAs in intra- and extracranial tumors may facilitate our ongoing study on the dual-targeting pan-anticancer strategy OncoCiDia for brain tumors [20,21,22].Given the dynamically altering genetic profiles in cancer cells, choosing a treatment modality that only targets tumor stromal cells deems to be more clinically relevant for a comparison between intracranial and extracranial settings. CA4P, fosbretabulin, is a classical VDA derived originally from African bush willow Combretum caffrum, which binds to tubulin and destroys the cytoskeleton in the endothelial cells of tumor blood vessels. Thus, based on orthotopic craniofacial tumor models of immunocompetent rodents [23], the present intraindividual comparative study was designed to test the follwoing working hypotheses: (1) a VDA may impose diverse efficacies between intracranial and extracranial tumors with the same origin but different locations; and (2) pre-treatment MRI may provide suggestive information for therapeutic prediction and patient selection. 4. Discussion

The current study demonstrated that (1) extracranial tumors showed better pre-treatment blood perfusion than intracranial tumors did, a phenomenon indicative of better drug penetration and/or distribution and, thus, therapeutic effect in favor of extracranial ones; (2) a better CA4P-induced vascular shutdown effect with a larger area of necrosis could be observed in the extracranial tumors, relative to intracranial lesions; and (3) such differential therapeutic efficacies could be reflected by a 3-fold concentration difference of CA Dotarem® (thus, CA4P of a comparable molecular weight) as defined by AUC300. To our knowledge, this is the first intraindividually comparative study for multiparametric MRI analyses between intracranial and extracranial tumors, to explore the role of a disrupted BBB in intratumoral anti-cancer drug penetration and retention, in particular with CA4P as a representative VDA.

Here, we adopted dynamic 3D multiparametric MRI, instead of merely 2D dimensional morphology MRI, for longitudinal follow-up based on VDAs’ unique mechanism for inducing necrosis [32]. Striking difference between intra- and extracranial tumor lesions could be dynamically detected by DCE imaging and the contrast agent concentration could be further quantified by AUC30 and AUC300, from which pharmacokinetics of small molecular anti-cancer drugs with short plasma half-life such as VDAs could be extrapolated. These measures reflect a comprehensive presence of perfusion, diffusion, permeability, vascular volume, and microenvironment exemplified by disrupted BBB that interplay with a particular drug such as CA4P. These parameters may represent potential biomarkers to select patients who may benefit from VDA treatment. The endpoint defined at 24 h after VDA treatment for this study was due to the considered optimal addition to the patient on this time point with a necrosis-avid radioactive tracer for OncoCiDia strategy [17,18,19].Despite the more homogeneous characteristics of tumor stroma, a different microenvironment also impacts the performance of VDAs, showing in this study with brain tumors as an extreme example. Intraindividual comparison between liver and pancreatic tumors of the same origin showed better response of liver cancer to CA4P, which could be attributed to the multiple arterial supplies to pancreatic tumors, which counteracted VDA’s vascular shutdown effect [33]. Similarly, an interindividual study comparing the therapeutic response to a different VDA DMXAA between ectopic (subcutaneous) and orthotopic (intramuscular) tumors showed a higher vascular volume in orthotopic tumors but a better vascular shutdown in ectopic tumors [34]. Another study on DMXAA comparing the therapeutic effect between intracranial and subcutaneous lesions indicated the marginal response in intracranial lesions [35].The tumor microenvironment is crucial for angiogenesis through cross-talks between a network of tumor cells, stromal cells, and endothelial cells [36]. To better mimic the microenvironment, especially the presence of immune cells which are believed to be associated with angiogenesis, an immune-competent model was preferred [37]. The intra- and extracranial VDA efficacy in immune-compromised mice showed that the concentration, measured by LC–MS/MS, was 25-fold higher in subcutaneous tumors and, thus, more massive necrosis was observed [31]. However, the different perfusion characteristic between subcutaneous and orthotopic tumors was inconsistent: the orthotopic hormone-sensitive prostate tumor model showed poorer perfusion compared with subcutaneous tumors [38]. All the studies above, together with our findings, highlight the crucial role of the tumor microenvironment in stroma-targeting strategy. For further translational studies on especially perfusion-related therapy, pre-treatment MRI characterization could be predictive for cancer control and thus needed for the selection of patients who might benefit from VDAs.

Regarding the marginal therapeutic effect intracranially, there are two possibilities: (1) limited drug penetration or distribution; and (2) intrinsic unresponsiveness of the intracranial tumoral endothelium to VDAs. Based on the findings here, limited CA concentration in the brain caused by the BBB may help explain the poorer therapeutic effect. Further studies elaborating on the combination of focused ultrasound or other methods that help increase the local distribution or retention of anti-cancer drugs in the brain may better confirm the findings here. However, this is out of the scope of the current study. If the inferior therapeutic effect persists after fostered intracranial drug penetration and retention, another possibility would be the intrinsic unresponsiveness, which can be clarified from biological aspects via bioinformatics such as single-cell sequencing and metabolic profiling.

The BBB is a barrier maintaining brain physiological hemostasis, by which transportation across the BBB by both transcellular and paracellular pathways are strictly controlled. Under certain circumstances such as stroke, Alzheimer’s disease, and multiple sclerosis, remodeling of the inter-endothelial protein complex may lead to BBB breakdown [39]. During carcinogenesis, the BBB is abrupted with leaky characteristics in the tumor area. Both BBBs comprise the vascular environment of homing tumor cells [40]. Despite an initially leaky BBB in the brain, the duration and magnitude for the BBB opening remain largely unknown. Furthermore, the initial leakiness may be insufficient for delivering the required drugs with high enough concentrations for cancer control. Further BBB opening can be achieved either physiologically or mechanically, and the major rationale is to transiently disrupt the BBB by decreasing the expression of TJ protein such as claudin-1, occluding, and tricellulin. Example measures include high osmotic pressure induced by intra-arterial mannitol [41], cereport (a bradykinin analog) [42], and borneol [43]. Mechanically, focused ultrasound together with microbubbles disrupts the BBB by reducing the expression of tight junction protein, resulting in a temporary opening [44]. Practically, properly optimized DCE-MRI can evaluate the subtle BBB leakage with moderate-to-excellent reproducibility and thus serves as a powerful tool for facilitating development of novel BBB-opening agents [45]. Besides the increasing influx of anti-cancer agents into BM, the inhibition of efflux from the BBB mediated by transporters encoded by the ATP-binding cassette gene family, such as P-glycoprotein, would also be an alternative method or a method that can be combined with [25]. These transporters are proven to be functional even when the BBB integrity is disrupted [46].This study was subject to the following limitations. Firstly, due to limited intracranial space and lower resolution of DWI, the comparison of ADC derived from the 3.0 T clinical magnet (Supplementary Figure S2B) was impaired by the partial volume effect. Secondly, in our study, measuring the concertation of VDAs was impossible since (1) the biological half-life for CA4P is around 25 min, and degradation happens at 24 h after treatment and during sampling; and (2) the loss of blood during sampling may lead to measurement bias of the VDA concentration. Thirdly, the AUC300 adopted here as a surrogate for drug perfusion should be interpreted with caution, due to the different molecular weight, charge distribution, and so on.