A major issue for the effective treatment of brain metastasis is the late stage of diagnosis with existing clinical tools. The aim of this study was to evaluate the potential of vascular cell adhesion molecule 1 (VCAM-1)-targeted MRI for early detection of brain micrometastases in mouse models across multiple primary tumor types.
Experimental Design: Xenograft models of brain micrometastasis for human breast carcinoma (MDA231Br-GFP), lung adenocarcinoma (SEBTA-001), and melanoma (H1_DL2) were established via intracardiac injection in mice. Animals (n = 5–6/group) were injected intravenously with VCAM-1–targeted microparticles of iron oxide (VCAM-MPIO) and, subsequently, underwent T2*-weighted MRI. Control groups of naïve mice injected with VCAM-MPIO and tumor-bearing mice injected with nontargeting IgG-MPIO were included.
All models showed disseminated micrometastases in the brain, together with endothelial VCAM-1 upregulation across the time course. T2*-weighted MRI of all tumor-bearing mice injected with VCAM-MPIO showed significantly more signal hypointensities (P < 0.001; two-sided) than control cohorts, despite a lack of blood–brain barrier (BBB) impairment. Specific MPIO binding to VCAM-1–positive tumor-associated vessels was confirmed histologically. VCAM-1 expression was demonstrated in human brain metastasis samples, across all three primary tumor types.
VCAM-1–targeted MRI enables the detection of brain micrometastases from the three primary tumor types known to cause the majority of clinical cases. These findings represent an important step forward in the development of a broadly applicable and clinically relevant imaging technique for early diagnosis of brain metastasis, with significant implications for improved patient survival.
This article is featured in Highlights of This Issue, p. 451
Advancing treatment options for brain metastases reinforces the need for detection earlier than can be achieved with existing diagnostic tools. Noninvasive imaging, specifically MRI, remains the mainstay of current clinical diagnostic methods in brain metastasis detection. In this study, we show that endothelial vascular cell adhesion molecule 1 (VCAM-1) is a relevant biomarker through early upregulated expression in the development of brain micrometastases from breast cancer, lung cancer, and melanoma; the three primary tumor types with the highest propensity to metastasize to the brain. Moreover, similar upregulation of VCAM-1 is present in human brain metastasis samples from all three primary tumor types. By targeting a surrogate biomarker, independent of blood–brain barrier (BBB) breakdown, MRI has been enhanced to enable detection of brain metastases in mouse models across a selection of primary tumors, at a considerably earlier stage than is possible with current, clinically used methods such as gadolinium contrast enhancement.
Cerebral metastasis, the spread of malignant tumors from an extracranial primary site of origin to the brain, is a leading cause of cancer mortality and morbidity (1). Collectively, secondary cancers are the most common intracranial malignancy, with non–small cell lung cancer (NSCLC), breast cancer, and melanoma demonstrating particular inclination to spread to the brain and contributing up to 4/5 clinical presentations (2). Patients with brain metastases typically survive for less than 6 months from diagnosis, unless definitive treatment proves possible (3). The poor survival is attributed in large part to the late stage of diagnosis and failure of current diagnostic methods to detect occult micrometastatic disease (4, 5).
Gadolinium contrast–enhanced MRI remains the preferred clinical diagnostic method for detecting brain metastases. In the context of neuro-oncological imaging, this modality offers far superior soft tissue contrast and anatomic characterization than other imaging techniques, such as CT and PET (6, 7). However, it relies on contrast agent extravasation in the presence of a disrupted blood–brain barrier (BBB), which is not usually seen in early-stage tumors. Thus, expanding the potential of current imaging modalities for detecting occult brain micrometastases (<2 mm diameter; ref. 8) could open up the diagnostic window. With the development of targeted agents that have improved BBB penetration, such as tesevatinib (9) and osimertinib (10), and new strategies for selective BBB permeabilization at metastatic sites (11), a wider range of therapeutic options for patients with brain metastases are emerging, particularly for microscopic disease where intervention will be most effective.
Metastatic hematogenous dissemination is a multistep process, for which extravasation of tumor cells into the distant organ is a key step. Specific cell surface proteins, including those belonging to the family of cell adhesion molecules (CAM), have been shown to facilitate the binding between tumor cells and the endothelium. We have previously demonstrated that, in breast cancer brain metastasis mouse models, vascular cell adhesion molecule 1 (VCAM-1) expression is upregulated early in micrometastasis-associated blood vessels and remains upregulated throughout metastatic seeding to, and colonization of, the brain. Furthermore, by conjugating anti-VCAM-1 antibodies to microparticles of iron oxide (MPIO), cerebral micrometastases could be detected using MRI before visibility with conventional gadolinium-enhanced MRI (12). However, breast cancer is not the only primary tumor with a propensity to metastasize to the brain, and both lung adenocarcinoma and melanoma contribute substantially to the incidence of brain metastasis. Consequently, with a view to determining the potential for clinical translation of this method, we sought to determine whether VCAM-targeted MRI is more broadly applicable for early micrometastasis detection.
The primary aims of this study were (i) to determine whether VCAM-1 is upregulated early in tumor development in xenograft mouse models of lung adenocarcinoma and melanoma brain metastasis, and (ii) to determine whether VCAM-1–targeted MRI enables detection of micrometastases prior to evidence of BBB breakdown independent of primary tumor type.
Materials and Methods
Three human-derived cell lines were used in this study: MDA231Br-GFP cells (subclone of metastatic breast carcinoma that preferentially metastasizes to the brain; kind gift from Prof. P. Steeg, National Cancer Institute, Bethesda, MD), H1_DL2 cells (brain metastasis–derived melanoma; kind gift from Prof. F. Thorsen, University of Bergen, Bergen, Norway), and SEBTA-001 cells (brain metastasis-derived lung adenocarcinoma; kind gift from Prof. G. Pilkington, University of Portsmouth, Portsmouth, United Kingdom). Following resuscitation from liquid nitrogen storage, no cell line underwent more than five passages prior to in vivo injection. MDA231Br-GFP cells were maintained in DMEM (Sigma-Aldrich) supplemented with 10% FCS (Thermo Fisher Scientific) and 1% l-glutamine (Life Technologies) in a 5% CO2 atmosphere at 37°C. H1_DL2 cells were maintained in DMEM supplemented with 10% FCS, 2% l-glutamine and 16 mL nonessential amino acid (100×, Sigma-Aldrich) in a 5% CO2 atmosphere at 37°C. SEBTA-001 cells were maintained in DMEM supplemented with 2% human serum (Sigma-Aldrich) in a 5% CO2 atmosphere at 37°C. All cell lines in active culture underwent routine Mycoplasma testing every fortnight, following the recommended protocol of the MycoAlert Mycoplasma Detection Kit (Lonza).
Female SCID Balb/c variant mice (8–9 weeks old; 19 ± 0.8 g; Charles River Laboratories) were anaesthetized with 2%–3% (vol/vol) vaporised isofluorane in oxygen and injected in the left cardiac ventricle under ultrasound guidance (Vevo 3100 Imaging System; Fujifilm VisualSonics), with 1 × 105 MDA231Br-GFP, H1_DL2 cells or SEBTA-001 cells in 100 μL PBS, as described previously (11, 12). To establish tumor progression over time, animals (n = 3 per time point) were sacrificed and brains harvested for histologic examination at weekly intervals (up to 28 days) for MDA231Br-GFP and H1_DL2 tumor-bearing mice, or at fortnightly intervals (up to 56 days) for SEBTA-001 tumor-bearing mice. Animals undergoing MRI were imaged at either 21 days (MDA231Br-GFP and H1_DL2) or 42 days (SEBTA-001) after intracardiac injection. Imaging time points were selected on the basis of the growth rates of the different cell lines, such that a time point was chosen for each model at which the metastases were established within the brain, but still within the micrometastatic phase prior to BBB breakdown. Owing to their relatively slower growth pattern, the SEBTA-001 cells were imaged at a later time point than the MDA231Br-GFP and H1_DL2 cells.
All animal experiments were approved by the University of Oxford Clinical Medicine Ethics Review Committee and the UK Home Office [Animals (Scientific Procedures) Act 1986], and conducted in accordance with the University of Oxford Policy on the Use of Animals in Scientific Research, the ARRIVE Guidelines, and Guidelines for the Welfare and Use of Animals in Cancer Research (13).
In vivo imaging
VCAM-MPIO or control IgG-MPIO was injected intravenously into each animal prior to MRI, as described previously (12). See Supplementary Materials and Methods for full details of antibody conjugated MPIO synthesis.
MRI data were acquired using a 7.0T MRI spectrometer (Agilent Technologies Inc.). On the day of imaging, tumor-injected mice were anesthetized with 2%–3% (vol/vol) vaporised isofluorane in 70% nitrogen:30% oxygen and injected intravenously via a tail vein with 4 mg Fe/kg body mass VCAM-MPIO (n = 5–6 per group) or IgG-MPIO (n = 4–5 per group) in 100 μL saline. A further cohort of naïve SCID mice were injected intravenously with VCAM-MPIO as above (n = 5). At 30 minutes after MPIO injection, animals were positioned in a customized cradle inside a quadrature birdcage coil (26 mm internal diameter; RAPID Biomedical GmbH). Respiration monitoring was performed and body temperature was maintained at approximately 37°C.
Prior to image acquisition for each animal, the main magnetic field (B0) inhomogeneity was corrected by active shimming. A preimaging scan was performed for each animal, to warm up the spectrometer to a stable running temperature, as follows: Multi gradient-echo three-dimensional (3D; MGE3D) sequence, flip angle = 15°, repetition time (TR) = 65.1 ms, echo time (TE) = 2.5 ms, 2nd echo time (TE2) = 4.0 ms, number of echoes (NE) = 15, spectral width (SW) = 150 kHz, averages (NT) = 1, matrix size = 256 × 192 × 96, field of view (FOV) = 22.5 × 22.5 × 22.5 mm, and total warming up time approximately 20 minutes. For VCAM-MPIO detection, a T2*-weighted 3D gradient echo dataset (MGE3D) was acquired as above, except matrix size = 256 × 192 × 192 and total acquisition time approximately 40 minutes. The mid-point of acquisition was 1 ± 0.2 hours after MPIO injection. Data were zero-filled to 256 × 256 × 256, to a final isotropic resolution of 88 μm, and the final images were reconstructed offline by adding individual echoes using the square root of a sum-of-squares algorithm. Subsequently, a set of ten coronal T1-weighted images (slice thickness = 1 mm) was acquired using a two-dimensional (2D) spin-echo sequence [TR = 500 ms, TE = 11.5 ms, matrix size = 128 × 128, number of slices (NS) = 10, NT = 1, FOV = 25 × 25 mm], both pre- and 5 minutes postintravenous gadolinium-DTPA (Omniscan; GE Healthcare) injection (30 μL), to identify BBB permeability.
The combined image of the multiple gradient echoes was constructed by using the square root of the sum of squares of signal intensities taken on a pixel-by-pixel basis from the individual echoes, as described previously (14). Each dataset of sum-of-squares images acquired by T2*-weighted imaging was manually masked and segmented to exclude extracerebral structures using ITK-SNAP (itksnap.org). Automated image processing of segmented images were performed using a custom designed MATLAB code (15). Briefly, hypointense signals were defined as a voxel value 0.65 times less than the mean value. Signals arising from ventricles or sinuses, which appear hypointense naturally, were excluded by imposing an upper threshold size limit of 20 voxels. A lower threshold filter of 1 voxel size was used to exclude noise. The threshold cut-offs and automated analysis were optimized in prior work to enable a detection rate of 98.3 ± 0.49% of total brain hypointensities (15). Segmented images were reconstructed to visualize the spatial distribution of MPIO binding, with hypointense voxels assigned to the red channel (Supplementary Fig. S1). Voxel volumes were summed and expressed as raw volumes in microliters.
IHC for VCAM-1 expression
Brain tissue sections from tumor-bearing mice and control cohorts were assessed immunohistochemically for VCAM-1 expression and additionally examined for colocalisation to metastasis-associated vasculature. Selected sections of human brain metastases for breast cancer, lung adenocarcinoma, and melanoma, obtained via image-guided biopsy (Walton Research Tissue Bank reference: 11/WNo03/02), were assessed immunohistochemically for VCAM-1 upregulation. See Supplementary Materials and Methods for full details of tissue sampling and IHC analysis.
For MRI hypointensity and tumor volumes, differences between animal cohorts for all tumor types were identified by one-way ANOVA. Post hoc Tukey tests were used to identify specific differences between groups. VCAM-1–positive staining was calculated on the basis of the number of strong intensity positive pixels in proportion to the brain volume. Difference in VCAM-1 expression across multiple time points was assessed by one-way ANOVA tests for each tumor type, with post hoc Tukey tests. All statistical analyses were two-sided and performed in GraphPad Prism (v.7; GraphPad Software).
Time course of metastatic growth
All animals reached their predefined endpoint without overt clinical signs. Metastases were present in all mice following tumor cell injection and were disseminated throughout the brain parenchyma. Metastatic growth varied according to the tumor type (Fig. 1A–C) over the experimental time course. Overall tumor burden in mice injected with SEBTA-001 cells was substantially lower than that of mice injected with either MDA231Br-GFP or H1_DL2 cells over the first 28 days from inoculation (Fig. 1D–F), indicating a much slower rate of growth in this phase. At the selected imaging time points for each tumor type, the mean ± SD tumor burden for MDA231Br-GFP was 0.033 ± 0.013 μL (21 days), 0.0076 ± 0.0008 μL for H1_DL2 (21 days), and 0.027 ± 0.019 μL for SEBTA-001 (42 days). VCAM-1 expression in proximity to metastatic colonies for all tumor types was upregulated from the earliest time point (Fig. 2A–F) and maintained throughout the time course. Colocalization of VCAM-1 specifically along the endothelial lining was confirmed by immunofluorescence (Fig. 2G–I).
In some cases, VCAM-1 expression appeared not to colocalize with the endothelium, likely reflecting expression on the surface of an out-of-plane tumor or glial cell; VCAM-1 is known to be expressed on some tumor cells (16), as well as microglia and astrocytes (17). It should be noted that VCAM-1 upregulation on other cells is not relevant to the methodology presented here, because the VCAM-MPIO remain intravascular and, thus, only have access to luminal VCAM-1 on endothelial cells.
In vivo detection of VCAM-1 upregulation by MRI
Marked hypointensities were evident on T2*-weighted images for all tumor-bearing mice injected with VCAM-MPIO (Fig. 3A–C). In comparison, few hypointensities were observed on T2*-weighted images from tumor-bearing mice injected with nontargeting IgG-MPIO or naïve animals injected with VCAM-MPIO (Supplementary Fig. S2). Histologic examination confirmed the presence of widespread micrometastases for all tumor types. Spatial correlation between MRI detectable hypointensities and the presence of metastases was assessed. Owing to differences in spatial resolution between MRI and histology (MGE3D isotropic resolution of 88 μm3 vs. maximum histologic slice thickness of 10 μm), it was necessary to stack alternate histologic sections within each image slice space to generate a composite image containing all metastases within that imaging slice (Fig. 3D–F). IHC analysis demonstrated VCAM-1–positive vessels in close proximity to micrometastases (Fig. 3G–I) and the presence of VCAM-MPIO within VCAM-1–positive vessels (Fig. 3J–L).
Quantitatively, significantly greater volumes of hypointensities were found in VCAM-MPIO–injected tumor-bearing animals than the IgG-MPIO control cohorts (Fig. 3M). In contrast, no significant differences were evident between naïve animals injected with VCAM-MPIO or IgG-MPIO.
For all mice injected with either MDA231Br-GFP or H1_DL2 cells, no contrast enhancement was evident on postgadolinium T1-weighted images (Fig. 4). One mouse injected with SEBTA-001 cells did show one discrete area of gadolinium enhancement, in the left cerebral cortex, on postcontrast T1-weighted imaging (Supplementary Fig. S3). This region correlated with a large cerebral metastasis (volume = 0.12 μL) confirmed histologically. This metastasis was also visible on T2*-weighted MRI following VCAM-MPIO administration, alongside additional focal hypointensities where gadolinium enhancement was not evident. This nonconforming metastasis was excluded from the subsequent hypointensity analysis to avoid detracting from our objective of detecting micrometastatic disease (i.e., prior to the size threshold required to produce contrast enhancement).
Assessment of VCAM-1 expression and VCAM-MPIO MRI positivity
Individual brain metastases across all mice were classified according to IHC detection of VCAM-1 on vessels in proximity to each metastasis. The mean ± SD for the number of brain metastases found for each tumor type were as follows: 102 ± 30, 22 ± 6, and 12 ± 5 tumors per animal for MDA231Br-GFP, H1_DL2, and SEBTA-001, respectively. Despite the lower number of SEBTA-001 tumors, as demonstrated in Fig. 5, some of the individual tumors were considerably larger than either the MDA231Br-GFP or H1_DL2 metastases. In some cases, VCAM-1 expression was observed up to 150 μm from the nearest metastasis.
As shown in Fig. 5A–C, the majority of tumors were associated with upregulated VCAM-1. For MDA231Br-GFP metastases, overall 89% were VCAM-1 positive, and for tumors above the median volume (6.5 × 10−4 μL), 98% were VCAM-1–positive. Similarly, overall 72% of H1_DL2 metastases were VCAM-1 positive, with 84% of tumors greater than the median volume (1.4 × 10−4 μL) associated with VCAM-1 expression. For the SEBTA-001 tumors, 93% elicited VCAM-1 upregulation on IHC and in the cohort of tumors above the median volume (2 × 10−4 μL), 96% were VCAM-1–positive. Even below the median volume, the majority of MDA231Br-GFP and SEBTA-001 metastases were VCAM-1–positive rather than negative (80% vs. 20% and 89% vs. 11%, respectively). In contrast, below the median volume the percentage of tumors that were VCAM-1–positive versus negative was closer to equal for the H1_DL2 group (60% vs. 40%).
Individual metastases were also correlated with the T2*-weighted MRI to determine the presence of corresponding hypointensities. Notably, because numerous vessels can be activated to be VCAM-1–positive around a tumor, this can give rise to more than one hypointense foci corresponding to a single metastasis (Supplementary Fig. S4). As shown in Fig. 5D–F, the majority of metastases corresponded to a hypointense signal on MRI, with 82%, 72%, and 89% positivity for the MDA231Br-GFP, H1_DL2, and SEBTA-001 tumors, respectively. As for VCAM-1 expression, the majority of tumors above the median volume were MRI positive; 94%, 92%, and 100% for the MDA231Br-GFP, H1_DL2, and SEBTA-001 tumors, respectively. Again, as for VCAM-1 expression, below the median volume, a greater percentage of MDA231BR-GFP and SEBTA-001 metastases were MRI-positive (69% vs. 31% and 78% vs. 22%, respectively), whereas numbers were approximately equal for the H1_DL2 group (52% vs. 48%).
VCAM-1 expression in human brain metastasis
IHC of image-guided brain metastasis biopsies from human brain showed prominent VCAM-1 expression in the brain parenchyma adjacent to the brain metastases from breast carcinoma, lung adenocarcinoma, and melanoma in all the cases studied (Fig. 6). The location of the biopsy specifically targeted the brain–tumor interface and along this edge, small metastatic foci were present, corresponding to early stages of invasion. VCAM-1 expression was upregulated along the tumor border and in close association with micrometastatic foci (Fig. 6A–C). VCAM-1 upregulation was evident predominantly on the endothelium in proximity to the tumor tissue (Fig. 6D–F).
Brain metastasis is an increasing clinical burden, as patients with cancer survive extracranial disease owing to improved systemic anticancer treatment (18, 19), and earlier diagnosis is critical. In this study, we have shown that a targeted contrast agent with specific binding to VCAM-1 permits detection of brain micrometastases below the limits that conventional clinical methods (passive gadolinium enhancement) allow. Importantly, using in vivo models of brain metastases for breast cancer, melanoma, and lung adenocarcinoma, we have demonstrated that VCAM-1–targeted MRI is applicable for the detection of micrometastases in the brain in multiple primary tumor types; together breast cancer, melanoma, and NSCLC, including lung adenocarcinoma, currently comprise almost 80% of the brain metastasis diagnoses.
Injection of tumor cells into the left cardiac ventricle under ultrasound guidance is a well-characterized method of inducing brain metastases in animal models (18). This approach successfully recapitulates the conditions under which tumor cells are hematogenously transported to a distant organ and results in widely disseminated synchronous brain metastases. Because the mouse brain is not equivalent to human brain, the pattern of dissemination observed throughout the brain may not be fully representative of human brain metastasis. However, it is known that metastases can present both cortically and subcortically in the human brain, and to this extent, the model reflects the human condition. Moreover, our data demonstrate that micrometastases can be detected anywhere within the brain using the VCAM-targeted approach.
The MDA231Br-GFP (subclone of metastatic human breast carcinoma) cell line and the H1_DL2 (human brain metastasis–derived melanoma) cell line have previously been shown to have specific tropism to the brain when injected intracardially in vivo (17, 20). The SEBTA-001 cell line is derived from a human brain metastasis of lung adenocarcinoma origin and has only been maintained in in vitro cell culture previously (21). This study has recapitulated previous studies with the MDA231Br-GFP and H1_DL2 cell lines, and has further demonstrated the ability of the SEBTA-001 cell line to induce brain metastases when injected intracardially in the mouse. In all the cases, an increase in tumor burden was evident over time, together with marked VCAM-1 upregulation from the earliest time points.
These results support the concept that targeting VCAM-1 is a good strategy for detecting brain micrometastases, with more than 70% of metastases showing upregulated expression selectively on nearby vasculature. Moreover, VCAM-1 expression in the micrometastatic stages was shown to be independent of primary tumor type, thus supporting the broad clinical applicability of this diagnostic approach. Importantly, as a protein present on the endothelial lining, VCAM-1 provides an attractive surrogate marker for early micrometastatic disease that exploits changes in the metastatic microenvironment, in spite of an undisrupted BBB.
The observation that VCAM-1–activated vessels are present outside the immediate confines of the tumor periphery (up to ∼150 μm from a nearby metastasis) means that a hypointensity focus does not give a precise location for the metastasis with which it is associated. This apparently distant activation may represent a locoregional inflammatory reaction stimulated by the tumor metastasis, for example through an immune-mediated cytokine release or indeed the presence of another metastasis out of plane. Consequently, local and focused therapy may not be possible on the basis of the VCAM-targeted approach. Nevertheless, this method has a major strength in assessing the overall metastatic burden within the brain and for informing on the application of systemic therapies that can target disseminated disease.
In one recent study, it was shown that adjuvant systemic treatment, with either targeted therapy or immunotherapy, improves the median overall survival by over 8 months for patients with melanoma brain metastases treated upfront with SRS for definitive local control (22). Thus, evidence is emerging that intervention with systemic therapies in microscopic disease will be most effective in the presymptomatic stage. With the arrival of new generation targeted drug therapies and immunotherapies, detection of micrometastatic disease may expand the therapeutic options for patients, leading to reduced morbidity and mortality.
The current imaging gold standard for detecting brain metastases is MRI with gadolinium contrast enhancement. The effectiveness of gadolinium-based contrast is limited to detecting metastases of sufficient size for the BBB to have become compromised; in experimental models, this typically occurs in tumors >500 μm in diameter (23), whereas human metastases must be 2–5 mm in diameter before they become visible by MRI (8). In this study, with the exception of one experimental subject with a large SEBTA-001 metastasis, no animals demonstrated contrast enhancement on T1-weighted MRI following intravenous gadolinium-DTPA injection.
Superparamagnetic iron oxide particles produce a contrast effect through distortion of the magnetic field, resulting in local field inhomogeneities that are detectable via T2*-weighted MRI (24). By targeting VCAM-1 with MPIO, we have shown that it is possible to detect brain micrometastases before they are visible on conventional gadolinium-enhanced MRI across in vivo models of breast cancer, melanoma, and lung adenocarcinoma. A significant proportion of tumors found histologically could be colocalized to a corresponding hypointense signal on MRI, and the percentages of MRI-positive metastases (82%, 72%, and 89% for breast, melanoma, and lung, respectively) and VCAM-1–positive metastases (89%, 72%, and 93%, respectively) were in close accord. The smallest detectable micrometastasis was approximately 50 μm in diameter. Improvements in the T2*-weighted MRI used in this study, from a gradient echo 3D (GE3D) sequence to a MGE3D sequence (25), has significantly increased contrast-to-noise ratios compared with earlier studies (12), thus increasing the sensitivity to signal hypointensities caused by MPIO.
In the larger metastases (>median volume), the majority of tumors were both VCAM-1 and MRI-positive in all models, supporting the sensitivity of this marker for early detection; it should be noted that these tumors are still considerably smaller (median tumor size ∼1 × 103 cells) than the current level of detection that is possible clinically (∼ I × 107 cells; ref. 12). Moreover, even below the median volume, the majority of metastases in the MDA231Br-GFP and SEBTA-001 models were both VCAM-1- and MRI-positive, whereas approximately 50% of the H1_DL2 tumors were associated with endothelial VCAM-1 expression and corresponding MRI detection. Thus, although there is likely to be a size limit below which a VCAM-1–targeted approach is less reliable, these tumors will still become detectable using VCAM-1–targeted MRI much earlier than through conventional gadolinium-enhanced imaging.
Minimal contrast effects were evident in any of the control cohorts, indicating the high specificity of the VCAM-1–targeted approach. Variation in the hypointensity volume for naïve animals injected with VCAM-MPIO may be accounted for by low levels of nonspecific VCAM-1 upregulation. However, no significant signal disruption compared with the other control cohorts was evident in these mice, indicating that any constitutive expression will not markedly reduce the sensitivity of VCAM-MPIO MRI for tumor detection.
IHC of coregistered image-guided biopsies of human brain metastasis confirmed the upregulation of VCAM-1 in tumor-associated blood vessels for all three main primary tumor types in all samples assessed. In particular, small metastatic foci present at the tumor edge representing the invasive edge of the tumor, were closely associated with VCAM-1–positive vessels. These invasive edge metastatic foci can be considered to be analogous to early micrometastatic disease (26), which is not readily accessible in human post-mortem tissue. Although the spatial resolution achievable in mice is considerably higher than is currently possible clinically we have extrapolated, on the basis of detection limits observed in nonclinical field strengths for mice and measurements from normal human brain using common clinical resolutions at 3T, that metastases of ≥300 μm in diameter should be detectable clinically with this approach (12).
Ideally, the chosen biomarker, VCAM-1, would only be present on vessels that are stimulated by proximity to malignancy. However, it is known that CAMs, including VCAM-1, maybe upregulated on the cerebral vasculature by a host of inflammatory diseases, for example multiple sclerosis (27), infection (28), and stroke (29). Nevertheless, where “at risk” patients with primary cancer are being assessed with this approach, the probability that intracranial VCAM-1 expression reflects metastatic involvement rather than an alternative pathology is greatly increased. At the same time, it is expected that the spatial presentation of VCAM-1 upregulation would be substantially different depending on the underlying cause and, therefore, the 3D information afforded by MRI would substantially offset the potential confounds of alternative diseases. Moreover, gadolinium contrast enhancement, the current clinical gold standard, is also not specific to malignancy; diagnostic uncertainty between a cerebral abscess, metastatic tumor, or primary malignancy is not uncommon for a solitary enhancing lesion on T1-weighted MRI (7).
The transition from bench-to-bedside is anticipated to lead to three key clinical improvements: (i) identification of discrete clinical problems where VCAM-targeted MRI will aid clinical decision making between the use of local versus systemic therapies; (ii) screening of high-risk populations to identify micrometastatic disease, which will inform intervention based upon existing systemic paradigms; and (iii) building sufficient lead time to allow testing of new therapeutic approaches specific to brain micrometastatic disease.
Targeted brain metastasis screening, for a defined “at risk” population of asymptomatic patients to identify subclinical disease, would represent a major paradigm shift in the management of patients with cancer. As an example, for patients with small cell lung cancer (SCLC), prophylactic cranial irradiation (PCI) not only reduces the risk of brain metastasis forming, but also improves both overall survival and disease-free survival where complete remission is achieved through systemic chemotherapy (30). However, there is a growing body of concern that some patients may experience unnecessary harm from adverse effects of PCI exposure, where there is minimal risk of future intracranial relapse (31). Detection of micrometastatic disease using VCAM-MPIO MRI would allow stratification of patients into a susceptible group likely to derive benefit from prophylactic treatment strategies and those who should remain under observation.
Early detection of substantial micrometastatic burden (that is currently undetectable) could eliminate costly treatment of both primary tumors and detectable brain metastases that is unlikely to result in significant benefit and, consequently, yield cost savings. A health economics impact assessment, conducted in conjunction with the NIHR-Diagnostic Evidence Cooperative, Imperial College London (London, United Kingdom), concluded that a diagnostic agent used for both diagnostic accuracy and disease monitoring in “at risk” patients would be cost effective based on a cost–utility approach measuring cost per quality-adjusted life years (personal communication; Dr Melody Ni, Imperial College London, London, United Kingdom).
Moreover, emerging strategies for drug delivery that overcome the BBB, for example through its permeabilization with tumor necrosis factor α (11), vasoactive substances such as bradykinin (32) or ultrasound-generated microbubbles (33) will expand the role of systemic agents for managing brain metastases. Coupling such approaches with a diagnostic method that widens the therapeutic window by detecting micrometastatic disease, which is more likely to respond to systemic agents, may significantly improve survival in the patients with lung cancer, breast cancer, and melanoma with subclinical brain metastases.
We have recently developed a nonimmunogenic, fully humanized anti–VCAM-1 antibody and biodegradable multimeric MPIO (mMPIO; ref. 34) to enhance the safety profile of the targeted contrast agent for human use. The mMPIO have the additional advantage of exhibiting approximately three times greater T2 relaxivities compared with commercially available polystyrene-coated MPIO; thus potentiating the contrast effect in human application (34). Following preclinical toxicology studies, it is anticipated that this agent will go forward to phase I/IIa clinical trial. The primary aim of this trial would be safety and pharmacokinetics, with initial assessment of efficacy as a secondary outcome. This early-phase clinical trial will focus on patients with known brain metastases, confirmed by standard MRI, and will follow a dose escalation design with expansion for dose-limiting toxicity, if encountered. Preliminary assessment of efficacy will be assessed in a subsequent expansion cohort, and the number of lesions identified with the VCAM-targeted approach compared with gadolinium-enhanced MRI will be determined. Following validation, recruitment for downstream clinical trials should focus on cancers with systemic therapies with proven or expected BBB penetration, as this will be the treatment of choice for patients with detectable micrometastatic disease; however, success at this stage should lead to rapid deployment of this technique to other patient groups.
In summary, VCAM-1–targeted MRI offers the opportunity for greater personalization of care in many existing treatment paradigms, with tangible benefits to both the individual and the healthcare service. With ongoing therapeutic advances and as the threat of brain metastases rises, in conjunction with increasing cancer survival, it can be envisaged that the cost–benefit ratio will tip in favor of wider surveillance measures such as those offered by this VCAM-1–targeted MRI approach. As a widely accessible technology in modern medicine with broad patient acceptability, MRI remains an ideal diagnostic tool for neuro-oncological imaging. Therefore, the minimal adaptations required for application of VCAM-1–targeted imaging to existing infrastructure will prove advantageous for its clinical translation.
Disclosure of Potential Conflicts of Interest
M.R. Middleton reports receiving speakers bureau honoraria from Bristol-Myers Squibb and UCB, and is a consultant/advisory board member for Immunocore, Rigontec, Agalimune, Novartis, and Roche. No potential conflicts of interest were disclosed by the other authors.
Conception and design: V.W.T. Cheng, M.S. Soto, M.R. Middleton, N.R. Sibson
Development of methodology: V.W.T. Cheng, M.S. Soto, A.A. Khrapitchev, F. Perez-Balderas, N.R. Sibson
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V.W.T. Cheng, M.S. Soto, A.A. Khrapitchev, F. Perez-Balderas, R. Zakaria, M.D. Jenkinson, M.R. Middleton, N.R. Sibson
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V.W.T. Cheng, M.R. Middleton, N.R. Sibson
Writing, review, and/or revision of the manuscript: V.W.T. Cheng, M.S. Soto, A.A. Khrapitchev, R. Zakaria, M.D. Jenkinson, M.R. Middleton, N.R. Sibson
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V.W.T. Cheng, R. Zakaria
Study supervision: M.R. Middleton, N.R. Sibson
This work was supported by a Cancer Research UK programme grant (C5255/A15935) to N.R. Sibson, M.S. Soto, A.A. Khrapitchev, and F. Perez-Balderas; and a Cancer Research UK/Medical Research Council Oxford Institute for Radiation Oncology–funded clinical research fellowship to V.W.T. Cheng. We gratefully acknowledge Prof Patricia S. Steeg (National Cancer Institute, Bethesda, MD), Prof Frits A. Thorsen (University of Bergen, Bergen, Norway), and Prof Geoffrey Pilkington (University of Portsmouth, Portsmouth, United Kingdom) for their kind gift of the human cell lines, MDA231Br-GFP, H1_DL2, and SEBTA-001, respectively, used in this study. We also gratefully acknowledge support from Mr Khaja Syed at the Walton Research Tissue Bank (Liverpool, United Kingdom).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.