Integrin-targeted quantitative optoacoustic imaging with MRI correlation for monitoring a BRAF/MEK inhibitor combination therapy in a murine model of human melanoma

Purpose To investigate αvβ3-integrin-targeted optoacoustic imaging and MRI for monitoring a BRAF/MEK inhibitor combination therapy in a murine model of human melanoma. Materials and methods Human BRAF V600E-positive melanoma xenograft (A375)-bearing Balb/c nude mice (n = 10) were imaged before (day 0) and after (day 7) a BRAF/MEK inhibitor combination therapy (encorafenib, 1.3 mg/kg/d; binimetinib, 0.6 mg/kg/d, n = 5) or placebo (n = 5), respectively. Optoacoustic imaging was performed on a preclinical system unenhanced and 5 h after i. v. injection of an αvβ3-integrin-targeted fluorescent probe. The αvβ3-integrin-specific tumor signal was derived by spectral unmixing. For morphology-based tumor response assessments, T2w MRI data sets were acquired on a clinical 3 Tesla scanner. The imaging results were validated by multiparametric immunohistochemistry (ß3 –integrin expression, CD31 –microvascular density, Ki-67 –proliferation). Results The αvβ3-integrin-specific tumor signal was significantly reduced under therapy, showing a unidirectional decline in all animals (from 7.98±2.22 to 1.67±1.30; p = 0.043). No significant signal change was observed in the control group (from 6.60±6.51 to 3.67±1.93; p = 0.500). Immunohistochemistry revealed a significantly lower integrin expression (ß3: 0.20±0.02 vs. 0.39±0.05; p = 0.008) and microvascular density (CD31: 119±15 vs. 292±49; p = 0.008) in the therapy group. Tumor volumes increased with no significant intergroup difference (therapy: +107±42 mm3; control +112±44mm3, p = 0.841). In vivo blocking studies with αvβ3-integrin antagonist cilengitide confirmed the target specificity of the fluorescent probe. Conclusions αvβ3-integrin-targeted optoacoustic imaging allowed for the early non-invasive monitoring of a BRAF/MEK inhibitor combination therapy in a murine model of human melanoma, adding molecular information on tumor receptor status to morphology-based tumor response criteria.


Introduction
Overactivation of the mitogen-activated protein kinase (MAPK) signal pathway by b-rapidly accelerated fibrosarcoma (BRAF) gene mutations V600E/K leads to uncontrolled proliferation of human cells and is a central mechanism of oncogenesis in melanoma [1,2]. Selective BRAF inhibitors (BRAFi) disrupt this oncogenic stimulus and demonstrate high initial tumor response rates in metastatic melanoma [3,4]. However, intrinsic or acquired BRAFi resistance limits long-term tumor response to BRAFi monotherapies [5]. One major mechanism of acquired BRAFi resistance is MAPK pathway activation by the mitogen-activated extracellular signal-regulated kinase (MEK), which may be overcome by selective MEK inhibitors (MEKi) [6]. Dual targeting of the MAPK signal pathway by a BRAFi/MEKi combination therapy demonstrated significantly improved overall and progression-free survival in patients with advanced BRAF-mutant melanoma compared to BRAFi monotherapy [7]. BRAFi/MEKi combination therapy is a first-line option in patients with BRAF-mutant metastatic melanoma (National Comprehensive Cancer Network Guidelines Version 1.2017, www.nccn.org).
Targeted therapies yield only subtle effects on tumor size and therefore limit the applicability of morphology-based criteria of tumor response [8]. As functional and molecular imaging allow for the non-invasive characterization of the tumor microenvironment beyond morphology, they bear the potential to provide novel, complementary imaging biomarkers of tumor response [9]. Functional and molecular imaging biomarkers may be better suited than sizebased response criteria in correlation with clinical endpoints such as early therapy response or progression-free survival [10,11].
α v ß 3 -integrin is a transmembrane protein overexpressed on angiogenic endothelium and tumor cells [12]. Depending on the investigated tumor model, α v ß 3 -integrin is a target structure for the non-invasive in vivo investigation of tumor angiogenesis and tumor cell populations [13]. In melanoma, α v ß 3 -integrin plays an important role in neoangiogenesis and tumor progression from the non-invasive, radial to the invasive, vertical growth phase [14,15]. Herzog et al. demonstrated the applicability of optoacoustic imaging with a targeted fluorescent probe for the characterization of α v ß 3 -integrin receptor status in human tumor xenografts in vivo [16].
The purpose of this experimental proof-of-principle study was to investigate α v ß 3 -integrintargeted optoacoustic imaging and MRI for the non-invasive in vivo monitoring of a BRAFi/ MEKi combination therapy in a murine xenograft model of human melanoma, validated by multiparametric ex vivo immunohistochemistry. We hypothesized that the α v ß 3 -integrin-specific optoacoustic signal would be significantly reduced under targeted therapy, delivering a surrogate of suppressed tumor α v ß 3 -integrin expression over the treatment course and adding quantitative, dual time point molecular information on the tumor microenvironment to morphology-based assessments of tumor response.

Materials and methods
The study was approved by the Government of Upper Bavaria Committee of Animal Research (Gz. ROB-55.2-2532.Vet_02-  and conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. All applicable institutional and/ or national guidelines for the care and use of animals were followed. Every effort was taken to reduce animal suffering. We kept the mice in individually ventilated cages (n = 4 mice per cage, relative air humidity 65% at n = 18 room air changes/h, temperature 26˚C, light-darkcycle 12 h), nourished with water and small animal nutrition. Nest boxes and nestles ensured environmental enrichment. Daily animal monitoring including weighing and tumor growth measurement was conducted. Abnormal inactivity was considered an indicator of pain and was treated by analgesia (0.5 mg/kg buprenorphine s. c.). The experiments were performed under isoflurane anesthesia (2.5% in 1.0 L of 100% O 2 per min for induction and 2.5% in 1.0 L of 100% O 2 per min for maintenance). Humane endpoints leading to euthanization were: maximum tumor diameter >1.5 cm, tumor exulceration, weight loss >15%, apathy, defense reaction when palpating tumors, respiratory problems, paresis, non-physiological body posture.

Optoacoustic imaging
Optoacoustic imaging was performed on a dedicated small animal imaging system (inVision 256-TF, iThera Medical GmbH, Munich, Germany) [17]. Briefly, a tunable optical parametric oscillator pumped by a neodymium-doped yttrium aluminum garnet laser provided excitation pulses with a duration of 9 ns at wavelengths from 680 nm to 980 nm at a repetition rate of 10 Hz with a wavelength tuning speed of 10 ms and a peak pulse energy of 100 mJ at 730 nm. Ten arms of a fiber bundle provided even illumination of a ring-shaped light strip of approximately 8 mm width. For ultrasound detection, 256 toroidally focused ultrasound transducers with a center frequency of 5 MHz (60% bandwidth), organized in a concave array of 270˚angular coverage and a radius of curvature of 4 cm, were used. Animals were scanned at the following n = 11 wavelengths: 700 nm, 730 nm, 740 nm, 750 nm, 760 nm, 770 nm, 780 nm, 790 nm, 800 nm, 850 nm, 900 nm. N = 10 pulses were averaged per wavelength, giving a temporal resolution per multispectral cycle of 11 s. After animals were placed into the system, the positions of the rostral and caudal ends of the tumor were identified to mark the start and stop locations of the scan (typically approximately 1 cm). A step size of 0.5 mm was used to scan through the tumor. Animals were scanned before and 5 h after i. v. injection of a commercially available α v ß 3 -integrin-targeted fluorescent probe (IntegriSense 750, 4 nmol; Perkin Elmer, Waltham, MA). The data was reconstructed with ViewMSOT (iThera Medical GmbH, Munich, Germany) using the back projection method. Reconstructed baseline images were used to model the optoacoustic profile of the tumor using an adaptive match filter spectral unmixing algorithm [18]. These models were then applied on a per mouse basis to the 5 h scans, where deviations from the model consistent with the absorption spectrum of the α v ß 3 -integrin-targeted fluorescent probe were revealed. Regions of interest (ROI) were manually drawn on single wavelength images to identify the tumor, and the average spectrally unmixed α v ß 3 -integrintargeted optoacoustic signal was quantified (arbitrary units, a. u.). Single wavelength background images were displayed in greyscale with a Frangi filter applied to enhance anatomical detail [19].

MRI
MRI was performed on a clinical 3 Tesla scanner (MAGNETOM Skyra, Siemens Healthineers, Erlangen, Germany). Animals were scanned head-first in prone position. T2w data sets were acquired using a 2D Turbo Spin Echo sequence (TR = 5470 ms, TE = 91 ms, in-plane resolution 0.3 x 0.3 mm, matrix size 192 x 192, slice thickness 1.5 mm). MR volumetry was conducted on an external workstation using dedicated post-processing software written in-house (PMI; Platform for Research in Medical Imaging, version 0.4) [20].

Planar whole-animal cryofluorescence imaging
In order to visualize the biodistribution of the targeted fluorescent probe, planar cryofluorescence images were acquired as described elsewhere [21]. Briefly, a cryostat was retrofitted with a light source, filters, and a camera which enabled acquisition of color images as well as fluorescence images of frozen specimen.

In vivo blocking studies
Competitive in vivo blocking studies were performed to confirm the target specificity of the α v ß 3 -integrin-targeted fluorescent probe. For blocking, α v ß 3 -integrin receptor antagonist cilengitide (800 μg; Selleck Chemicals, Houston, TX) was injected i. v. 15 min prior to injection of the α v ß 3 -integrin-targeted fluorescent probe (4 nmol). Optoacoustic imaging and signal quantification were then performed as described above and signal intensities of blocked and unblocked animals were compared accordingly.

Statistical analysis
Statistical analysis was performed using commercially available statistics software (SPSS 24, IBM Corp., Armonk, NY). The continuous parameters were expressed as means with standard deviations (95% confidence intervals). For intragroup comparisons (follow-up vs. baseline), the Wilcoxon test was applied, while the Mann-Whitney U test was applied for intergroup comparisons. The confirmatory tests (α v ß 3 -integrin-specific optoacoustic signal at follow-up vs. baseline; ß 3 -integrin in the therapy vs. the control group) were compared against Bonferroni-Holm-adjusted α' levels (α = 0.05). All other tests were compared against α = 0.05 as they were considered exploratory.

Optoacoustic imaging
The α v ß 3 -integrin-specific optoacoustic signal was significantly reduced under therapy, demonstrating a unidirectional decline in all animals (mean signal: from 7.98±2.22 a. u. to 1.67 ±1.30 a. u.; p = 0.043). No statistically significant change of the optoacoustic signal was observed in the control group, with a heterogeneous development of individual values (mean signal: from 6.60±6.51 a. u. to 3.67±1.93 a. u.; p = 0.500). There was no statistically significant difference in baseline optoacoustic signals between the therapy and the control group (mean signal: therapy 7.98±2.22 a. u. vs. control 6.60±6.51 a. u.; p = 0.690). Individual optoacoustic signal intensities at baseline and follow-up are displayed in Table 1 and Fig 1. Fig 2 shows a single wavelength optoacoustic image for anatomical reference in greyscale and the spectrally unmixed, α v ß 3 -integrin-specific signal in fire. Representative color-coded tumor maps of exemplary animals from the therapy and the control group before and after treatment are provided in Fig 3. Competitive blocking studies confirmed the target specificity of the α v ß 3 -integrin-targeted fluorescent probe (Fig 4). The α v ß 3 -integrin-targeted fluorescent probe

MRI
Tumor volumes increased in both the therapy and the control group with no statistically significant intergroup difference (ΔVol Therapy +107±42 mm 3 vs. ΔVol Control +112±44 mm 3 ; p = 0.841). There was no difference in baseline tumor volumes between the therapy and the control group (Vol TherapyBL 35±10 mm 3 vs. Vol ControlBL 62±33 mm 3 ; p = 0.151).
Individual tumor volumes at baseline and follow-up are provided in Table 1

Immunohistochemistry
In line with the optoacoustic imaging results, ß 3 -integrin expression was statistically significantly lower in the therapy group (ß 3 : 0.20±0.02 vs. 0.39±0.05; p = 0.008). We observed a statistically significantly lower microvascular density in the therapy group compared to the control group (CD31: 119±15 vs. 292±49; p = 0.008). There was a statistically non-significant lower tumor cell proliferation in treated compared to untreated animals (Ki-67: 3,925±1,693 vs. 5,782±1,092 p = 0.151). Individual values for the immunohistochemical parameters are provided in Table 2. Immunohistochemical stainings of representative tumor sections are provided in Fig 8.  adding complementary molecular information on α v ß 3 -integrin receptor status to morphology-based assessments of tumor response. Several groups investigated the potential of preclinical optoacoustic imaging with targeted probes for the non-invasive characterization of the tumor microenvironment in vivo [16,[22][23][24][25]. α v ß 3 -integrin-targeted fluorescent probes were studied to qualitatively visualize experimental human breast carcinomas and glioblastomas [16,24]. However, reproducible signal quantifiability is an essential prerequisite for therapy monitoring. Our study adds to the literature as it presents a first approach to quantify the fluorescent signal at different time points under therapy validated by immunohistochemistry, evaluating the potential of optoacoustic  imaging for therapy monitoring in future clinical trials. Optoacoustic imaging yields high translational potential with a focus on tumors located close to the body surface. Clinical optoacoustic imaging systems were already investigated in melanoma, inflammatory bowel disease, breast cancer, and thyroid cancer [26][27][28][29]. Clinical optoacoustic imaging allows for the acquisition of anatomical and functional information on tumors and healthy tissues exploiting intrinsic tissue contrast and visualizing endogenous absorbers (melanin and oxygenated/deoxygenated hemoglobin) [26,27]. However, in order to generate molecular information such as  tumor receptor status, exogenous absorbers, i. e., targeted probes, are necessary [30]. Despite the promising preclinical results reported in the literature, data on clinical optoacoustic imaging with targeted fluorescent probes are still lacking and data on the use of targeted fluorescent probes in humans are limited. Indocyanine green and methylene blue are two non-targeted near-infrared range fluorochromes approved by the Food and Drug Administration for the use in humans, but these unspecific blood pool agents fail to provide information on tumorinherent target structures [31]. In clinical studies, targeted fluorescent agents were already investigated for the guidance of surgical and endoscopic procedures [32][33][34]. For instance, a folate-receptor α-targeted fluorescent probe emitting in the visible spectrum (500 nm) was recently investigated in patients with ovarian or breast cancer to intraoperatively detect additional tumor lesions and to aid the definition of tumor margins [33]. Contrast agents that absorb in near-infrared, which would enable the deepest possible imaging in tissue for optoacoustics, are also available and were applied in clinical studies for optical imaging. For example, bevacizumab-IRDye800CW was used to detect colorectal peritoneal metastases intraoperatively with the use of a fluorescence camera [35]. Optoacoustic imaging could be used for similar applications, though its capability of deeper tissue penetration could potentially offer additional benefits intraoperatively or in non-invasive settings. However, biodistribution and biocompatibility of targeted, near-infrared fluorescence imaging agents in humans remain major issues that need to be addressed before clinical translation. Binding to non-target tissues as well as the enhanced permeability and retention effect may lower the target-tobackground signal and therefore limit the applicability for therapy monitoring [31,36]. For local staging under therapy, subcutaneous injection of the targeted probe analogously to interstitial MR lymphography with consecutive lymphatic transport to the target tumor (melanoma) would be a possible approach to reduce the systemic side effects in humans and may be investigated in future studies.
The role of optoacoustic imaging compared to alternative α v ß 3 -integrin-targeted imaging modalities remains to be defined. Positron emission tomography (PET) with targeted radionuclides as well as MRI using targeted contrast agents also allow for the non-invasive characterization of α v ß 3 -integrin receptor status in vivo [13,[37][38][39]. Compared to optoacoustic imaging, these modalities are not limited by low tissue penetration depth and are superior with regard Integrin-targeted optoacoustic imaging to whole-body tumor staging. However, the lack of ionizing radiation, lower costs compared to PET and MRI, and the availability of clinical imaging devices for bedside use are major advantages of optoacoustic imaging. One potential application of handheld optoacoustic scanners may be the monitoring of superficially-located target lesions, i. e., lymph node metastases, cutaneous metastases, or the primary tumor (if unresectable). This could be performed as part of a multimodality imaging protocol complementary to established imaging modalities such as 18 F-fluorodeoxyglucose-PET/computed tomography or MRI, generating a surrogate of wholebody tumor receptor status under targeted therapy. Integrin-targeted optoacoustic imaging

Limitations
We acknowledge several limitations of the present study. First, the animal number was small and the results including reproducibility of the quantitative data need to be validated in larger cohorts. However, this can be considered appropriate for the study purpose, which was to provide a first proof-of-principle of quantitative dual time point optoacoustic imaging for therapy monitoring in a small animal model of human melanoma. In addition, Joseph et al. recently provided evidence for the reproducibility and repeatability of quantitative in vivo optoacoustic imaging [40]. Second, the imaging results were validated by single time point immunohistochemistry and the follow-up interval was limited to one week. Additional follow-up intervals for both imaging and immunohistochemistry may reveal deeper insights into the development of α v ß 3 -integrin expression and the according optoacoustic signal over the therapy course. Third, α v ß 3 -integrin-targeted imaging may be limited by elevated background signal as α v ß 3integrin is not only expressed by tumors, but also several non-target tissues including the gut and the liver [41]. The localization of the tumor may thus have an effect on ROI selection and signal quantifiability. Nevertheless, the detected α v ß 3 -integrin-specific optoacoustic signal allowed for a good depiction of the tumor against the background in all animals, which can be attributed to the relatively high α v ß 3 -integrin expression of the A375 cell line compared to other available melanoma models [42].

Conclusions
In this experimental study, α v ß 3 -integrin-targeted optoacoustic imaging allowed for the in vivo monitoring of a BRAFi/MEKi combination therapy in a murine model of human melanoma. Our results provide a proof-of-principle of quantitative dual time point optoacoustic imaging for therapy monitoring in vivo.