Avoiding hitting a nerve
Fluorescence-guided surgery uses contrast agents to identify nerves intraoperatively; however, existing agents have poor tissue specificity and high background or permit only surface-level imaging. Wang et al. developed near-infrared nerve-binding fluorophores that could be administered locally or systemically for deep nerve visualization. Testing the oxazine-based contrast agents in murine and porcine models identified buried nerves during laparoscopic surgery. These contrast agents could help avoid injuring nerves during surgery.
Abstract
Nerve-binding fluorophores with near-infrared (NIR; 650 to 900 nm) emission could reduce iatrogenic nerve injury rates by providing surgeons precise, real-time visualization of the peripheral nervous system. Unfortunately, current systemically administered nerve contrast agents predominantly emit at visible wavelengths and show nonspecific uptake in surrounding tissues such as adipose, muscle, and facia, thus limiting detection to surgically exposed surface-level nerves. Here, a focused NIR fluorophore library was synthesized and screened through multi-tiered optical and pharmacological assays to identify nerve-binding fluorophore candidates for clinical translation. NIR nerve probes enabled micrometer-scale nerve visualization at the greatest reported tissue depths (~2 to 3 mm), a feat unachievable with previous visibly emissive contrast agents. Laparoscopic fluorescent surgical navigation delineated deep lumbar and iliac nerves in swine, most of which were invisible in conventional white-light endoscopy. Critically, NIR oxazines generated contrast against all key surgical tissue classes (muscle, adipose, vasculature, and fascia) with nerve signal-to-background ratios ranging from ~2 (2- to 3-mm depth) to 25 (exposed nerve). Clinical translation of NIR nerve-specific agents will substantially reduce comorbidities associated with surgical nerve damage.
INTRODUCTION
Iatrogenic nerve injury presents one of the most feared surgical complications and a major source of morbidity across all surgical specialties (1). Surgery incurs up to 600,000 nerve injuries annually in the United States alone (2), accounting for ~17% of all peripheral nerve trauma (3). Nerve damage occurs through varied insults with rates stratified by anatomic locations and surgical procedures, revealing an endemic problem across surgery (4, 5). Chronic neuropathies resulting from iatrogenic injury limit patient quality of life and increase healthcare costs (4). Although experienced surgeons readily identify major nerves under circumstances of normal anatomy, previous trauma, radiotherapy, tumors, and prior surgery can lead to fibrotic tissue deposition and atypical surgical dissection planes, thus making nerve delineation challenging. Minimally invasive surgery places peripheral nerves at further risk given reliance on assumptions about anatomical landmarks rather than on direct nerve visualization. In addition, some peripheral nerves, such as the delicate cavernous nerves responsible for continence and potency in men—often injured during prostatectomy—are so small that they prove difficult to recognize even under optical magnification (6). Electromyography, ultrasound, optical coherence tomography, and confocal endomicroscopy have been used to aid in intraoperative nerve identification (7–9). However, these techniques lack specificity, resolution, and wide-field imaging functionality, making real-time nerve detection difficult.
Fluorescence-guided surgery (FGS) has successfully integrated into clinical medicine with two near-infrared (NIR) U.S. Food and Drug Administration (FDA)–approved contrast agents [indocyanine green (ICG) and methylene blue] and has a bright future as a plethora of cancer-targeted molecular imaging agents navigating clinical trials enter surgical practice (10–16). Clinical open and laparoscopic FGS vision systems operate almost exclusively at NIR wavelengths (650 to 900 nm) (17–24), where tissue chromophore absorbance, autofluorescence, and photon scattering fall to local minima. The NIR spectral range permits low-resolution imaging up to 1 cm in depth, with micrometer-scale resolution at ~1- to 2-mm tissue depths (25, 26). Methylene blue and ICG, both blood pool (intravascular contrast) agents with minimal tissue-specific uptake, have demonstrated the clinical efficacy of FGS by lowering a diverse set of complications including normal tissue injury to bile ducts and ureters and inadequate perfusion in the case of anastomotic leaks (21, 27). Multiple NIR cancer-targeted agents are in the clinical translation pipeline with emission maxima spanning the NIR window (LUM015, λemission = 650 nm; OTL38, λemission = 800 nm) (10–16). Although management of cancer remains vital to decrease mortality, iatrogenic nerve injury prevention has arguably equal importance in preventing postsurgical morbidity.
Eight classes of nerve- or brain-specific fluorophores [stilbene, coumarin, styryl pyridinium, distyrylbenzene (DSB), tricarbocyanine, nerve-specific peptides, sodium channel selective peptides, and oxazine fluorophores] have been found, yet all nerve-specific contrast agents fluoresce at visible wavelengths and are plagued with high nonspecific tissue uptake (adipose and skin) arising from requisite lipophilicity (25, 28–42). These dual limitations produce characteristically shallow, surface-level tissue imaging depths (<100 μm) with nerve fluorescence intensities on par with nonspecific uptake in adipose, skin, and cut muscle edges. With the exception of nerve-targeted peptides, all classes of nerve-specific contrast agents consist of small-molecule organic fluorophores with low molecular weights (<500 Da) as required to cross the blood-nerve barrier (BNB) (43). Because of enhanced optical bioimaging metrics above 650 nm, NIR small-molecule fluorophores with favorable nerve-binding pharmacology offer the potential to generate high signal-to-background ratios (SBRs) for nerve visualization in deep, native tissue environments (31, 44). Surveying existing nerve-binding small-molecule fluorophore classes revealed select chemical scaffolds as promising base compounds for tuning fluorophore emission into the NIR range. Specifically, the oxazine scaffold, with nerve-specific derivatives currently emitting at visible wavelengths, offers a potentially promising, flexible architecture as close structural analogs display NIR emission yet entirely lack nerve specificity (35). Although oxazines have facilitated imaging in the NIR window to a degree from a dwindling emission shoulder reaching past 650 nm, most photons are emitted at visible wavelengths. Directed oxazine synthetic engineering offers the potential to identify translational small-molecule nerve contrast candidates capable of successfully delineating nerves in complex surgical environments.
We built an extensive fluorophore library through purposeful synthetic modification of the oxazine fluorophore scaffold, resulting in identification of contrast agents that are both highly nerve-specific and NIR-emitting. All oxazine derivatives (n = 64) underwent multi-tiered screening for NIR optical properties and nerve specificity. Nerve specificity evaluation involved nerve-to-muscle (N/M) and nerve-to-adipose (N/A) tissue SBR quantification, the latter an underreported or poorly optimized metric among other nerve-specific probes (SBR generally ≤1) (33, 40, 45). Potential lead candidate library compounds yielded impressive imaging metrics after direct and systemic administration in mice. The salient advantage of NIR nerve imaging agents was conclusively demonstrated by detection of ~200-μm-wide nerves beneath 2 mm of tissue, whereas fluorescent features were obscured or unresolved using visible nerve dyes. Potential lead NIR candidates enabled nerve imaging in robotic, minimally invasive surgery in pigs in which exposed iliac plexus nerves demonstrated SBRs > 20. LGW01-08 fluorescence displayed SBR ~2 at ~2- to 3-mm tissue depths in lumbar nerves beneath the intact body wall 20 min after intravenous injection, whereas conventional white-light endoscopy failed to delineate deep nerves. Critically, LGW01-08 and LGW05-75 produced strong nerve contrast against all key surgical tissue classes. The synthesized NIR nerve-specific fluorophores facilitated nerve imaging at depth, extending nerve detection from the tissue surface through 3 mm of intact native tissue.
RESULTS
Rational design of nerve-specific NIR oxazines
Oxazines 1 and 4 served as base compounds in the systematic development of NIR nerve-specific fluorophores, with the underlying hypothesis that similar structural properties should yield similar in vivo functionality. Oxazine 4 has bright in vivo nerve-specific fluorescence but maximum emission (635 nm) at visible wavelengths. Oxazine 1 falls within the NIR region (maximum emission = 680 nm); yet, both direct and systemic administration yield pan-tissue fluorescence without nerve specificity (35). Systematic modifications of the nerve-specific oxazine scaffold resulted in a 64-compound fluorophore library (Fig. 1A). The resulting derivatives fell into three structurally similar subgroups: (i) asymmetric oxazine 4 hybrids, (ii) asymmetric oxazine 1 hybrids, and (iii) symmetric oxazine fluorophores (fig. S1). See the Supplementary Materials for full synthetic details and structural characterization (table S1 and figs. S2 to S6). The fluorophore library’s maximum absorbance and emission values ranged from 590 to 701 nm and 606 to 718 nm, respectively, resulting in 42 compounds with NIR emission with varying brightness (Fig. 1B and table S2). All oxazine library derivatives had molecular weights ranging from 340.3 to 428.6 g/mol, below the theoretical threshold for BNB penetration (500 g/mol) (46, 47).
(A) Generalized chemical structures of the oxazine fluorophore library, representing all 66 oxazine derivatives, including oxazines 1 and 4. (B) Average nerve fluorescence intensity after systemic administration versus the maximum emission wavelengths for the entirety of the oxazine library. All fluorophores on the right side of the red dotted line had emission maxima in the NIR wavelengths (>650 nm). The average control nerve autofluorescence from uninjected mice is plotted as a horizontal black line, with dotted lines representing ±1 SD from the mean. The four potential lead NIR nerve-specific candidate oxazine derivatives selected for additional study, along with two base compounds oxazine 1 and 4, are color-coded as follows: LGW01-08 (teal), LGW05-75 (green), LGW04-31 (orange), LGW03-76 (red), oxazine 1 (purple), and oxazine 4 (blue). A.U., arbitrary units. The average N/M SBR calculated using the nerve and muscle tissue fluorescence intensities after (C) direct and (D) systemic administration of each oxazine derivative compared to oxazine 1 (Ox-1, purple), oxazine 4 (Ox-4, blue), and an unstained/uninjected control (CTRL, white) group. The oxazine fluorophores in each graph are organized by maximum emission wavelength in ascending order, where all fluorophores above the red dotted line had emission maxima in the NIR wavelengths (>650 nm). Average N/M ratios were calculated from data collected for n = 6 nerve sites per fluorophore and are presented as mean ± SD. Positive nerve-specific fluorophores were determined as those with N/M and N/A SBR >1.5 and are marked with asterisks for statistical significance. Data for each positive nerve-specific fluorophore were compared to control unstained/uninjected data to test for the significance of N/M SBR using a one-way ANOVA followed by a Fisher’s LSD multiple comparison test with no assumption of sphericity using the Geisser-Greenhouse correction, where *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. The four NIR potential lead candidate oxazine derivatives selected for additional study, LGW01-08 (teal), LGW05-75 (green), LGW04-31 (orange), and LGW03-76 (red), are color-coded for ease of visualization. All oxazine derivatives are labeled using their number only without their LGW prefix to aid in visualization.
Oxazine derivative library nerve specificity screening
All 64 oxazine derivatives were screened using predominately in vivo methods because ex vivo assays showed poor predictive power for in vivo nerve specificity (fig. S7). Consequently, all library compounds underwent both direct and systemic in vivo administration to identify potential lead candidates. Direct administration investigated nerve tissue binding specificity and generalized fluorophore-nerve interactions, whereas systemic injections selected for suitable pharmacokinetics and BNB penetration.
After both administration routes, brachial plexus– and sciatic nerve–containing regions were imaged to quantify nerve, muscle, and adipose tissue fluorescence (figs. S8, A to C, and S9, A to C). Absolute nerve fluorescence intensities varied widely across the library, with brighter overall signal after direct (fig. S8A) compared to systemic administration (Fig. 1B and fig. S9A). Muscle and adipose tissue intensity varied similarly after direct administration (fig. S8, B and C), whereas after systemic administration, more compounds yielded muscle and adipose fluorescence near endogenous autofluorescence thresholds (fig. S9, B and C). Measured tissue fluorescence intensities enabled N/M and N/A SBR quantification (Fig. 1, C and D, and figs. S8D and S9D). Mean N/M ratios ranged from 0.73 to 5.29 after direct administration (Fig. 1C) and from 0.69 to 3.43 after systemic administration (Fig. 1D). Mean N/A ratios spanned from 0.69 to 4.97 after direct administration (fig. S8D) and from 0.81 to 2.94 after systemic administration (fig. S9D).
Using N/M and N/A SBRs > 1.5 as criteria for positive nerve specificity, 39 and 14 oxazine derivatives yielded positive nerve contrast after direct and systemic administration, respectively. As expected, many compounds displayed positive nerve specificity via direct administration but were negative for nerve contrast after the more stringent systemic administration screening due to the molecules’ inherent unfavorable pharmacokinetics or low BNB penetration properties. Four oxazine derivatives (LGW01-08, LGW05-75, LGW04-31, and LGW03-76) were selected for further study on the basis of their superior NIR spectral properties, bright fluorescent signal, high nerve specificity, and structural diversity (Fig. 1, C and D, and fig. S10). LGW01-08, LGW05-75, LGW04-31, and LGW03-76 showed high nerve-to-background tissue contrast after both direct and systemic administration (Fig. 2, A and B). Although tissue-specific fluorescent intensities were variable, all four NIR oxazines displayed significantly higher N/M and N/A ratios than control (unstained) tissue after either direct (N/M: P = 0.0002 for oxazine 4, P = 0.0010 for LGW01-08, P = 0.0438 for LGW05-75, P = 0.0021 for LGW04-31, and P = 0.0013 for LGW03-76; N/A: P = 0.0225 for oxazine 1, P = 0.0054 for oxazine 4, P = 0.0039 for LGW01-08, P = 0.0449 for LGW05-75, P = 0.0119 for LGW04-31, and P = 0.0015 for LGW03-76) or systemic (N/M: P < 0.0001 for oxazine 4, P = 0.0002 for LGW01-08, P < 0.0001 for LGW05-75, P = 0.0015 for LGW04-31, and P = 0.0016 for LGW03-76; N/A: P = 0.0307 for oxazine 1, P < 0.0001 for oxazine 4, P = 0.0002 for LGW01-08, P = 0.0029 for LGW05-75, P = 0.0030 for LGW04-31, and P = 0.0012 for LGW03-76) administration (Fig. 2, C to F). Furthermore, biodistribution studies revealed that all four NIR nerve-specific compounds showed the highest normalized fluorescence in nerve tissue over any other tissue or organ (fig. S11). LGW01-08 and LGW05-75 were selected for penetration depth and large-animal imaging studies due to high absolute nerve fluorescence intensity and SBRs (Fig. 2, D and F).
Chemical structures, representative photographs, and fluorescence images of NIR oxazine derivatives LGW01-08, LGW05-75, LGW04-31, and LGW03-76; oxazine 1 and oxazine 4; and an unstained control group after (A) direct application (125 μM) to exposed sciatic nerves and (B) systemic administration at 200 nmol/mouse. All images are representative of data collected for n = 6 nerve sites per fluorophore. Intensity metrics shown in boxes of each fluorescence image denote the fraction of the brightest fluorophore’s intensity that the image represents between images in the same row. Scale bar, 3 mm. Ex, maximum excitation. Em, maximum emission (nm). Mean nerve, muscle, and adipose tissue fluorescence intensities are displayed for all fluorophores and control groups after (C) direct and (D) systemic administration as in (A) and (B). Quantified nerve SBRs were calculated for comparison between potential lead compounds after (E) direct and (F) systemic administration as in (A) and (B). All quantified data are presented as the mean ± SD. Data for each fluorophore were compared to the unstained control group to test for nerve SBR significance using a one-way ANOVA followed by a Fisher’s LSD multiple comparison test with no assumption of sphericity using the Geisser-Greenhouse correction, where *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001.
NIR LGW01-08 increases nerve visualization depths
To quantify bioimaging improvements derived from NIR nerve contrast agents, known thicknesses of mouse muscle tissue were used as a controlled experimental method for studying nerve imaging as a function of tissue depth. Murine brachial plexus nerves were stained via systemic administration of the visible DSB derivative 4,4′-[(2-methoxy-1,4-phenylene)di-(1E)-2,1-ethenediyl]bis-benzenamine (BMB; maximum emission = 505 nm) or via direct administration of visible rhodamine B [a nerve-specific rhodamine analog of oxazine 1, maximum emission = 570 nm (48)], visible oxazine 4, or NIR LGW01-08 (Fig. 3). Direct administration provided uniform fluorescence intensity across both small and large diameter nerves, enabling quantitative tissue depth studies of variable-sized nerves at similar fluorescence intensities, whereas systemic administration was necessary to generate BMB nerve-specific contrast. Image acquisition was completed for oxazine 4 and LGW01-08 using excitation and emission filters analogous to the 700-nm NIR fluorescence channel in clinical FGS systems. Filters tuned to their respective imaging channels were used for rhodamine B and BMB fluorescence imaging (23, 49, 50).
(A) Photographs and fluorescence images of the mouse brachial plexus after systemic administration of BMB (1 μmol/mouse, Ex/Em: 393/503 nm) or direct administration of rhodamine B (125 μM, Ex/Em: 550/580 nm), oxazine 4 (125 μM, Ex/Em: 620/640 nm), or LGW01-08 (125 μM, Ex/Em: 640/660 nm). Fluorescence images were collected on the exposed nerves and with 1 to 2 mm of mouse muscle tissue layered atop the surgically exposed brachial plexus region. Yellow arrows indicate regions of high nonspecific adipose or cut muscle uptake. Blue arrows mark large nerves visible to 1- to 2-mm tissue depths. Blue dotted lines indicate overlaid muscle tissue edge boundaries. Red dashed line and arrow denote the location and direction, respectively, of the parallel-spaced, micrometer-scale nerve line profiles measured in (B). Orange dashed line and arrow denote the location and direction, respectively, of the brachial plexus nerve line profile measured in (C). Scale bar, 3 mm. (B) Fluorescence intensity cross-sectional profiles for rhodamine B, oxazine 4, and LGW01-08 quantified for the highlighted small nerve branches at 1-mm [red dashed lines in (A)] and (C) the large brachial plexus nerves at 2-mm [orange dashed lines in (A)] tissue depths. The quantified profile from rhodamine B resulted from nonspecific uptake from cut muscle tissue [marked by yellow arrow in (A)]. Black arrows denote the line profile peaks corresponding to nerves. (D) Three-dimensional fluorescence intensity topographical maps of BMB and LGW01-08 after systemic administration generated from images at the brachial plexus region beneath 1 mm of muscle tissue. The color bar represents pixel intensity value in arbitrary units (A.U.). All data are representative of data collected for n = 3 nerve sites per fluorophore.
LGW01-08 demonstrated marked improvements in NIR imaging metrics at depth when compared to visible oxazine 4, which has only a dwindling NIR emission shoulder. First, LGW01-08 identified micrometer-scale (~200 μm) and large-diameter (~1 mm) nerves at up to 2-mm depths; oxazine 4 provided only large-diameter and limited micrometer-scale nerve identification beneath 1 mm of muscle tissue, with all features completely obscured by 2-mm tissue depths (Fig. 3A). Second, LGW01-08 displayed superior resolving power at millimeter tissue depths. Parallel micrometer-scale nerves spaced ~350 μm apart were selected for both oxazine-based dyes at the tissue surface level. Cross-sectional profiles at these spatial locations beneath 1 mm of tissue demonstrated that solely LGW01-08 could resolve both nerves as two well-defined Gaussian peaks. The more diffuse point-spread function resulting from nerve-specific visible oxazine 4 signal produced an unresolved single fluorescent peak beneath 1 mm of tissue (Fig. 3B). Last, LGW01-08 facilitated nerve detection at 2-mm tissue depths, whereas oxazine 4 could not (Fig. 3C)—a clear benefit for iatrogenic nerve injury prevention in clinical environments.
LGW01-08 yielded an even larger performance disparity between other, more hypsochromically shifted, visible fluorophore scaffolds. For instance, rhodamine B provided nerve identification only on exposed surface nerves with no fluorescence detected at depth besides nonspecific cut muscle signal (Fig. 3, A to C). Similarly, BMB suffered from complete signal attenuation at a 1-mm depth due to high tissue chromophore absorbance at 505 nm along with high adipose tissue nonspecific uptake (Fig. 3, A and D). In addition, both rhodamine B and BMB yielded substantial nonspecific uptake in adipose tissue and cut muscle edges, with lower overall nerve specificity than the oxazine scaffold. Because BMB required intravenous injection for nerve accumulation, a comparison to intravenously injected visible oxazine 4 and NIR LGW01-08 was conducted, which yielded a similarly large performance boost for the NIR nerve-specific probe over visible nerve fluorophores (fig. S12). BMB demonstrated a near-uniform featureless surface intensity at 1-mm depths, whereas NIR LGW01-08 showed low background autofluorescence characteristic of NIR fluorophores, which aided in the visualization of a well-defined nerve structure beneath 1 mm of muscle (Fig. 3D). In addition, LGW01-08 was capable of identifying large-diameter nerves at 2-mm depths, structures that were again unresolvable via oxazine 4 fluorescence imaging (fig. S12). Thus, LGW01-08 resolved nerves at high fidelity across all imaging depths with negligible nonspecific uptake in muscle, adipose, and cut muscle edges, a feat not feasible with the hypsochromically shifted rhodamine B, BMB, or oxazine 4 due to suboptimal optical and pharmacologic properties.
Deep tissue NIR fluorescent nerve imaging in minimally invasive surgery
Systemic NIR LGW01-08 and LGW05-75 nerve probe administration in swine fluorescently highlighted peripheral nerves in the pelvic region during minimally invasive surgery on the da Vinci Si Surgical System (Fig. 4, A to C, and movies S1 and S2). Within 20 min after injection, LGW01-08 revealed bright lumbar nerve signal against the lateral and anterior body wall (movie S3). Buried nerve tissues, largely invisible in conventional white light, were unambiguously delineated with fluorescence (SBR ~1.5) underneath peritoneal tissue and several millimeters of mixed adipose and muscle tissues (Fig. 4, A and B; fig. S13A; and movies S4 and S5). Clinically relevant tissue types, including adipose, muscle, fascia, and vasculature, showed low nonspecific uptake enabling notable nerve detection (fig. S13B). Nerves on all tissue backgrounds generated positive contrast with varied underlying tissue background, resulting in a broad range of semiquantitative SBRs from ~1.5 in lumbar nerves running between fascia and muscle to ~25 in exposed iliac nerves running along the iliac bifurcation (Fig. 4, C and D; fig. S13, B and C; and movies S6 and S7). In addition, LGW01-08 exhibited strong nerve contrast on unexpectedly short time scales (~20 min) after systemic administration, the fastest reported time to achieve positive nerve contrast against muscle and adipose tissue by two- to six fold (33, 35, 40, 45).
(A) Photographs of conventional white-light illumination (first column), LGW01-08 (0.34 mg/kg, intravenously) deep lumbar nerve visualization at 2- to 3-mm tissue depths using NIR fluorescence (second column), and dissection under fluorescence guidance (third column). Gold standard H&E staining of the resected linear fluorescent structures is shown in the fourth column. (B) Photographs of conventional white-light illumination (first column), LGW05-75 (0.34 mg/kg, intravenously) deep lumbar nerve visualization at ~1-mm tissue depth using NIR fluorescence (second column), and dissection under fluorescence guidance (third column). Gold standard H&E staining of the resected linear fluorescent structures is shown in the fourth column. Black arrows denote the location of nerve tissue in the H&E image (scale bar, 500 μm). (C) Photographs of conventional white-light illumination (first and third columns) and LGW01-08 (0.34 mg/kg, intravenously, second column) and LGW05-75 (0.34 mg/kg, intravenously, fourth column) iliac nerve detection and visualization over varying tissue types. (D) Fluorescent intensity cross-sectional profiles of surgically exposed iliac nerves visualized with LGW01-08 (SBR ~1.15 to 1.2) and LGW05-75 (SBR ~5 to 25). Nerves and other tissue types are labeled according to the region of the profile they represent. Nerve SBRs were approximated from the highest nerve peak to the range of background tissue intensities. SBRs are reported as approximations due to the variable tissue working distances, varied fields of view, nonplanar tissue surfaces, and variable spatial distribution of excitation light. (E) Photographs of conventional white-light endoscopy and fluorescence images from mock surgical procedures performed in the presence of nerve tissue with LGW05-75. The blue arrow and dotted yellow lines in case 1 indicate a nerve-like structure seen in white light that showed no fluorescent signal and thus was unlikely to be a nerve. In case 2, no nerves were observed in white-light endoscopy, yet fluorescent imaging highlighted a contiguous sub–1-mm fluorescent nerve (blue arrow and dotted yellow lines) running between the surgical scissor blades in danger of transection. Yellow arrowheads in all panels denote nerve tissue locations in intraoperative images. The surgical instruments are outlined with red dotted lines. All data are representative of data collected for n = 3 swine per fluorophore.
A simulated fluorescence-guided dissection highlighted the benefit of NIR nerve imaging integration into minimally invasive surgery. In the first surgical case, nerve depth assessment using conventional white-light mode proved challenging. In addition, a white linear, potential nerve structure appeared near the instrument blades, adding visual confusion in the conventional white-light view. Fluorescent imaging provided distinct identification of the true nerve running below the fascial layer, thus avoiding risk from accidental transection and confirming the white superficial structure as nonnerve (Fig. 4E). In the second surgical case, white-light endoscopy failed to reveal a small nerve between the blades of the surgical tool. Switching to fluorescence mode revealed faint signal emanating from the sub–1-mm-diameter nerve (Fig. 4E and movie S8). These examples illustrate the utility of NIR nerve probes for revealing subsurface nerve tissue before exposure, aiding in prevention of iatrogenic injury.
DISCUSSION
Systematic oxazine fluorophore library synthesis yielded structure-inherent NIR nerve probes, achieving imaging depths of ~2 to 3 mm as expected in NIR bioimaging, yet previously unrealized with existing visibly emissive probe designs. Tissue optical properties diverge sharply near the visible-to-NIR spectral boundary (~650 nm), accounting for the vastly improved NIR oxazine derivative imaging performance in the biologically transparent optical window. Oxy- and deoxy-hemoglobin shows 3.4- and 2.6-fold lower extinction coefficients, respectively, in the 650- to 700-nm NIR region versus within the 600- to 650-nm visible range, allowing NIR oxazine derivative fluorescent photon propagation to occur with substantially fewer absorption events (51). Consequently, LGW01-08 demonstrated fourfold higher nerve fluorescent signal at 1 mm and identified nerves at 2-mm tissue depths, whereas increased scattering from visible probes produced a nearly homogeneous or nonexistent fluorescent signal at 1- to 2-mm tissue depths. Additionally advantageous, the absorbance profiles of the NIR nerve-specific oxazines (λexcitation = 640 to 655 nm) are compatible with the existing infrastructure of FDA-approved NIR FGS systems (17–24, 49, 50).
NIR oxazines selected for optimal combined optical and pharmacologic properties demonstrated salient advantages over existing nerve-specific probes. First, potential lead candidates displayed minimal adipose uptake, a common problem for nerve-specific probes arising from a requisite degree of lipophilicity for BNB partitioning. For instance, DSB fluorophores have high nonspecific uptake in adipose tissue, with reported fluorescence surpassing nerve tissue (30, 31). Similarly, nerve-targeted peptides offer a flexible moiety for fluorophore bioconjugation with N/M SBR surpassing 10 yet also experience high adipose and skin uptake as well as nonspecific labeling of cut muscle edges (33, 40). Although DSB and nerve-targeted peptides show high N/M contrast, the fluorescent signal from human adipose tissue would likely vastly overwhelm this expected weaker nerve signal for most surgical tasks. Furthermore, thin fluorescent adipose layers would obscure underlying nerves, and filamentous adipose muscle inclusions may produce false positives for these adipose accumulating probes. Second, NIR oxazine probes contiguously stain peripheral nerves, in contrast to styryl pyridinium dyes that selectively stain the dorsal nerve roots and trigeminal ganglia (32, 34). Last, stilbene, DSB, and coumarin nerve agents require 380- to 405-nm excitation sources with ~500- to 550-nm fluorescence emission, limiting achievable imaging depths to the tissue surface. LGW01-08 and LGW05-75 contain a distinct combination of optical and pharmacological profiles that provide positive nerve contrast against key surgical tissue classes encountered in both open and laparoscopic surgery. With positive SBRs ranging from ~1.5 in subperitoneal nerves beneath muscle to ~25 in exposed iliac nerves above vasculature in swine, NIR oxazines excel from low background autofluorescence coupled with minimal nonspecific uptake. Last, unlike most nerve dyes, NIR oxazine fluorophores show high SBRs after both direct and systemic administration to afford varied surgical site delivery routes with both rapid and long-term nerve visualization for compatibility with a wide variety of procedures.
Subsequent translational research avenues include formulation development for both direct and systemic administration routes. All lipophilic nerve contrast agents, including the agents developed in this work, require small percentages of dimethyl sulfoxide and/or Cremophor EL surfactant injection volume fractions to yield adequate solubility (28–30, 35). Potential clinical formulations for systemic administration of nerve-specific fluorophores include polyethylene glycol–based phospholipids and cyclodextrin-based excipients given their amphiphilic nature and successful performance in FDA-approved drug products (52–56). However, these liquid-phase formulations would not be ideal for use in the clinic via direct administration, which will require innovative bioengineering strategies to enable contrast agent delivery to uneven, highly angled tissue surfaces potentially covered in blood, saline, or serous fluids. Thermosensitive hydrogels offer a promising approach with application of a free-flowing solution at ambient temperatures that adheres on surgical tissue plane contact via sol-gelation transitions designed to occur at body temperature (57–59). In addition, the general nerve location must be known a priori for direct application intraoperatively, and further diffusion depth studies will also be required to quantify detectability thresholds. Nevertheless, direct administration offers the potential for high nerve contrast using only a microdose of nerve-specific fluorophore as defined by the FDA, which would facilitate clinical translation using the expedited exploratory investigational new drug (eIND) guidelines (28, 60).
Other notable preclinical translational milestones must include preclinical pharmacology and toxicology studies toward Investigational New Drug (IND) approval by the FDA for the more broadly applicable systemic administration of nerve-specific fluorophores. Exploratory central nervous system (CNS) safety pharmacology evaluations should be performed to determine if the nerve-targeting mechanism inhibits action potential transduction, a requisite development step likely to undergo careful FDA scrutiny. Exemplary neurotoxicity studies required for IND approval include a CNS safety pharmacology core battery such as the Functional Observational Battery or modified Irwin’s test to assess neurotoxicity (61, 62). Promisingly, no mice at any dose in the NIR oxazine library screening displayed signs of behavioral neurotoxicity. Preliminary results indicate systemic maximum tolerated doses of 6 to 30 mg/kg in mice with variability in No Observable Adverse Effect Levels (NOAEL) depending on specific scaffold functional groups. To mitigate potential downstream candidate failure in costly Good Laboratory Practices (GLP) studies, subsequent clinical translation steps will involve screening potential lead library candidates (LGW01-08, LGW05-75, LGW03-76, and LGW04-31) using in vitro electrophysiological methods (gold standard ion channel assays) along with standard dose-range finding studies to evaluate toxicity through blood chemistries and necropsy in rodents.
NIR oxazine human doses and trial design depend strongly on the administration route and intended clinical application. Systemically administered LGW01-08 murine doses yielded both optimal imaging SBRs and strong absolute nerve intensities at ~0.5 mg/kg [human equivalent dose (HED) of ~40 μg/kg]. Lower imaging doses down to 0.1 mg/kg display a minimal SBR reduction but dose-proportional absolute drop in nerve tissue intensity. More sensitive fluorescence endoscopic vision systems may enable progressive dose reductions to compensate for lower nerve brightness with a decreased total administered dose for even greater safety margins. Compared to systemic administration, topical delivery provides ~10-fold higher absolute nerve intensities with ~20-fold lower NIR oxazine doses, permitting microdosing (<100 μg) when scaled to the HED (~60 μg for NIR oxazines) (28). Thus, direct administration at the microdose scale offers innovative eIND clinical development pathways that substantially reduce the burden of preclinical GLP animal testing before human subject studies (14, 60). Carefully selected clinical study patient populations designed to evaluate direct administration efficacy while minimizing patient exposure may include nerve imaging in limbs with diabetic peripheral artery disease or peripheral vascular disease immediately before surgical amputation. Early-phase efficacy evaluations for either administration route will require fluorescent nerve ground-truth verification methods to ensure fluorescence-to-nerve correlation. Potential study designs may include imaging of undisputed ground-truth nerves observed in white-light endoscopy or intraoperatively tagging nerves in excised tissues for postoperative histology confirmation of imaging results.
In summary, the synthetic design, development, and preclinical characterization of a focused oxazine library identified strong NIR nerve-specific fluorophores with diverse physicochemical properties and nerve-binding characteristics as potential lead candidates for clinical translation. Potential lead candidates displayed minimal nonspecific uptake in all surgically relevant tissues, including adipose tissue, with the highest reported nerve-to-background ratios (~25) and the deep imaging depths (~2 to 3 mm). Swine imaging with LGW01-08 and LGW05-75 demonstrated the power of NIR nerve probes, enabling identification of buried nerves invisible under conventional white-light endoscopy using a modified clinical laparoscopic FGS system. Potential lead candidate NIR oxazine probes show promise for first-in-human studies after IND approval, enabling pharmacology and toxicology testing either via traditional routes for systemic administration or via the eIND pathway with directly administered microdoses (~60 μg). Last, oxazine structure-activity relationships unearthed through library compound screening has provided insight into nerve-binding fluorophore molecular design principles, offering a powerful means for nerve specificity management and a potential synthetic roadmap deeper into the infrared region (~800 nm).
MATERIALS AND METHODS
Study design
The overarching goal of this project was to develop a contrast agent in which a single probe could be used as the nerve-targeting agent and the fluorescent reporter, with emission in the NIR and compatibility with the existing fluorescence-guided surgery clinical infrastructure. A library of 64 oxazine-based derivatives was designed and synthesized to combine the NIR fluorescent profile of pan-tissue fluorescent oxazine 1 and the nerve-binding properties of visibly emissive oxazine 4. All synthesized fluorophores were characterized for structural information, including high-performance liquid chromatography–mass spectrometry (HPLC-MS) analysis and purity confirmation, and for optical properties, including maximum absorption, excitation and emission wavelength, extinction coefficient, quantum yield, as well as brightness. Compounds were then grouped into visibly emissive and NIR-emissive categories, and screened for in vivo nerve specificity following both direct and systemic administration strategies. Regions of interest (ROIs) for the nerve, muscle, and adipose tissues were selected on the color image (blind to the fluorescence image) to quantify tissue fluorescence intensity using custom MATLAB code. Quantified tissue intensities were then used to calculate N/M and N/A as an indication of nerve specificity. Screened candidate fluorophores with NIR spectral properties, high nerve specificity, and in vivo brightness were selected for additional studies, including nerve tissue imaging at depth and large-animal studies. Mock fluorescence-guided dissection using lead candidate NIR nerve-specific fluorophores in the presence of nerve tissue in swine was completed, highlighting the benefit of NIR nerve imaging integration into minimally invasive surgery. All replication information is indicated in the figure legends and in the respective sections of Materials and Methods. Synthetic details, structural information, and optical property characterization are provided in the Supplementary Materials. No data were excluded from the report on this study. Primary data are reported in data file S1.
Oxazine derivative library synthesis
Oxazine 4 perchlorate and oxazine 1 perchlorate were obtained from Exciton. All chemical building blocks were purchased from Sigma-Aldrich, Thermo Fisher Scientific, or TCI America. Unless otherwise indicated, all commercially available starting materials were used directly without further purification. The 64 oxazine derivatives were synthesized in two to eight steps depending on the availability of the starting materials. The required starting materials that were not commercially available were synthesized according to described or modified literature protocols (see the Supplementary Materials, general protocols A to H) (63–65). Overall, 21 commercially available aromatic amines were used to prepare 90 intermediates, affording 64 oxazine derivatives (fig. S1). Analytical thin-layer chromatography was performed on ready-to-use plates with silica gel 60 (32 to 63 μm, EMD Millipore). Reaction products were purified by flash column chromatography using silica gel (Sorbent Technologies) or using the Biotage Isolera Flash Chromatography System with SNAP Ultra cartridges (Biotage). Mass spectra were measured on an Agilent 6244 time-of-flight tandem liquid chromatography mass spectrometry instrument with a diode array detector VL+ (Agilent Technologies; fig. S2). The potential lead candidate structures were further characterized by one-dimensional (1D) 1H and 13C, as well as 2D 1H-1H COSY and 1H-13C HMBC nuclear magnetic resonance (NMR) spectroscopies (figs. S3 to S6) on a Bruker 400-MHz Avance II+ spectrometer (Bruker). See the Supplementary Materials for full synthetic details, purity, and optical property analysis.
Animals
Approval for the use of all animals in this study was obtained from the appropriate Institutional Animal Care and Use Committees. Male CD-1 mice weighing 22 to 24 g were purchased from Charles River Laboratories. Before surgery, animals were anaesthetized with ketamine (100 mg/kg) and xylazine (10 mg/kg) (Patterson Veterinary) administered intraperitoneally. At the completion of all studies, mice were euthanized using inhalation of carbon dioxide followed by cervical dislocation. Female swine (Pork Power) weighing 60 to 70 kg were anaesthetized with 4% isoflurane. At the completion of all studies, swine were euthanized by intravenous administration of Euthasol (Virbac).
In vivo nerve specificity screening of the oxazine derivative library
Each compound was screened for its tissue specificity using two in vivo methods, including a direct (28) and systemic administration strategy, where nerve contrast was examined in murine brachial plexus and sciatic nerves. The oxazine derivatives were solubilized for in vivo use in the previously described co-solvent formulation (31). For direct administration, the previously optimized staining procedure was used (28), which is described briefly as follows. The brachial plexus and sciatic nerves were surgically exposed by removal of overlaying adipose and muscle tissues. The oxazine compound was formulated at 125 μM in the co-solvent formulation, and 100 μl was incubated on the exposed brachial plexus or sciatic nerve sites, covering all tissues in the ROI (nerve, muscle, adipose, and fascia) for 5 min. The fluorophore-containing solution was removed, and the entire stained area was irrigated with saline nine times, followed by a 5-min incubation with blank formulation, and then irrigated with saline nine more times to remove any unbound fluorophore. Images of the entire stained region were acquired 30 min following completion of staining. Unstained nerve sites were used for all control images to quantify autofluorescence. Each oxazine derivative was screened in n = 3 mice or 6 nerve sites per fluorophore.
The optimal dose and pharmacokinetics of oxazine 4 (28, 35) were used to screen the oxazine derivatives by systemic administration, where 200 nmol of each compound in 100 μl of co-solvent formulation was administered intravenously 4 hours before imaging. This fluorophore administration to imaging window had been previously shown to provide the highest nerve-to-background tissue fluorescence for oxazine 4 and several other nerve-specific fluorophores (30, 31, 35). Uninjected animals were used for all control images to quantify autofluorescence. Each oxazine derivative was screened in n = 3 mice or 6 nerve sites per fluorophore with the two brachial plexus and two sciatic nerve sites being averaged together for a total of two replicates per animal, one of each nerve type.
Intraoperative fluorescence imaging systems
A custom-built small-animal imaging system capable of real-time color and fluorescence imaging was used to acquire in vivo rodent images (66). Briefly, the imaging system consisted of a QImaging EXi Blue monochrome camera for fluorescence detection with a removable Bayer filter for collection of co-registered color and fluorescence images. A PhotoFluor II light source was focused onto the surgical field through a liquid light guide and used unfiltered for white-light illumination. For fluorescence excitation, the PhotoFluor II was filtered with a 620 ± 30–nm bandpass excitation filter, yielding a power density of 15 mW/cm2 at the imaging surface. The resulting fluorescence was collected with a 700 ± 37.5–nm bandpass emission filter. All filters were obtained from Chroma Technology. Camera and light source positions did not vary throughout the course of all imaging studies, allowing quantitative comparison of in vivo fluorescence intensities. Camera exposure times ranged from 5 to 2000 ms for fluorescence image collection.
A custom-built laparoscopic imaging system also capable of real-time color and fluorescence imaging was used to acquire in vivo swine images and videos. The imaging system was integrated into the clinical-grade da Vinci Si Surgical System (Intuitive Surgical). The modification to the clinical-grade system consisted of a Necsel Neon 5-W 640-nm laser (Necsel) coupled to the da Vinci Si endoscope with a 642-nm StopLine single-notch blocking filter (Semrock). The single-notch blocking filter was placed in a clinical Si camera sterile adaptor between the rod lens endoscope and the Si camera head. Fluorescent signal acquisition occurred in the clinical Si white-light mode with the blocking filter, removing the excitation light and the fluorescent signal detected primarily on the red, green, blue (RGB) red-Bayer elements of a Si endoscope camera head. Laser power at the endoscope tip measured 800 mW, with the optical power losses occurring primarily at the laser fiber/light guide and the Si camera head/rod lens interfaces. Fluorescence and color videos were captured using the clinical Si Vision side cart at an exposure time of 2 ms. Videos were recorded on a clinical Panasonic high-definition serial digital interface recorder connected to the TilePro video out connections on the da Vinci Si Vision side cart. Video clips were edited in iMovie to crop their length and include text overlays. Screen captures of the in vivo video clips were collected and are presented as figures. Da Vinci Si imaging display algorithms apply nonlinear gamma and color corrections to the collected video, and thus, although image quantification is possible to report what would be visible SBR to surgeons intraoperatively, these values are only semiquantitative and could vary from what is observed in a more controlled system.
Nerve imaging at depth studies using a model oxazine derivative
Oxazine 4 and LGW01-08 were used to stain mouse brachial plexus nerves using both direct and systemic administration methods with the same dose (100 μl of 125 μM fluorophore-containing solution for direct administration and 200 nmol of fluorophore in 100-μl volume for systemic administration) and administration methods previously described. Rhodamine B (Sigma-Aldrich) was also used to stain mouse brachial plexus nerves via direct administration (100 μl of 125 μM fluorophore-containing solution). BMB was used to stain mouse brachial plexus nerves via systemic administration using a dose of 1 mg per mouse administered via tail vein injection and a 4-hour imaging time point as previously optimized (31). The ability to image the brachial plexus nerve at depth was assessed under varying thicknesses of mouse muscle tissue. Muscle tissue was resected from the peritoneal muscle layer and measured for thickness. One and two layers of resected peritoneal muscle tissue were placed over the stained nerves, providing 1- and 2-mm-thick homogeneous tissue through which the stained nerve tissue was imaged. The same muscle tissue was used to cover the stained nerves from all fluorophores with care taken to orient the muscle in the same way to ensure equivalent comparison. The imaging system configuration was modified to reflect the 700-nm channel in clinically relevant FGS systems for image collection of oxazine 4 and LGW01-08 (23, 49, 50), where the PhotoFluor II was filtered with a 650 ± 22.5–nm bandpass excitation filter. The resulting fluorescence was collected with a 720 ± 30–nm bandpass emission filter for image collection. For fluorescence excitation of rhodamine B, the PhotoFluor II was filtered with a 545 ± 12.5–nm bandpass excitation filter and the resulting fluorescence was collected with a 605 ± 35–nm bandpass emission filter for image collection. For fluorescence excitation of BMB, the PhotoFluor II was filtered with a 405 ± 20–nm bandpass excitation filter and the resulting fluorescence was collected with a 525 ± 25–nm bandpass emission filter for image collection. Camera exposure times ranged from 5 to 4000 ms for fluorescence image collection.
Swine nerve imaging studies of oxazine derivatives
LGW01-08 and LGW05-75 were administered via intravenous injection at a body surface area–scaled dose from mice for swine nerve imaging studies (0.34 mg/kg) (67). Nerves running along the body wall in the peritoneal cavity buried beneath 1 to 3 mm of fascia and adipose tissue were imaged using the previously described fluorescence-enabled laparoscopic imaging system docked to the da Vinci Si Surgical System. Nerve contrast was monitored immediately following injection and up to 6 hours after systemic administration. Dissection and resection of the identified nerve structures were performed using the da Vinci, where the resected tissues were fixed in formalin. The fixed tissues were sent to the Oregon Health & Science University Histopathology Shared Resource for paraffin embedding, sectioning, and subsequent gold standard hematoxylin and eosin (H&E) staining.
Image analysis of intraoperative nerve contrast
Custom-written MATLAB code was used to analyze the tissue-specific fluorescence, in which ROIs were selected on the nerve, muscle, and adipose tissues using the white-light images but blinded to the fluorescence images (doi: 10.5281/zenodo.3698316). These ROIs were then analyzed on the co-registered matched fluorescence images permitting assessment of the mean tissue intensities as well as the N/M and N/A ratios. Fluorescence intensity measurements were divided by the exposure time to obtain normalized intensity per second measurements. Mean nerve-to-background tissue ratios were calculated for each oxazine derivative. Positive nerve specificity was determined via the following criteria: mean N/M > 1.5 and mean N/A > 1.5. These criteria were chosen because oxazine 1, the non–nerve-specific parent fluorophore, demonstrated positive N/A ratios by statistical significance alone, and thus, the criteria were made more stringent to reduce the number of false positives identified as nerve specific. For nerve imaging at depth image analysis, fluorescent line profiles and 3D intensity topographical plots were generated in ImageJ using the plot profile and surface plot functions, respectively. Line profiles for nerve imaging at depth analysis were background- and exposure time–corrected. Line profiles and SBRs quantified from large-animal laparoscopic studies were assessed from the RGB images displayed on the da Vinci Surgical System, where no background of exposure time correction was applied. The tissue-level SBR may differ slightly from the displayed-level SBR as image-signal processing pipelines contain nonlinear elements, including gamma and color corrections that take place before video capture in a da Vinci Si endoscope system, and thus, SBRs are reported as semiquantitative. In addition, nonuniform tissue surfaces, variable endoscope-tissue positioning, and vignetting add further complications to resolving absolute tissue-level SBRs. The reported SBRs represent the nerve signal displayed to a surgeon on the da Vinci high-resolution stereo viewer and thus are representative of clinical integration and SBRs that would be visible to the surgeon.
Statistical analysis
Significant differences between nerve SBR means were evaluated using a one-way analysis of variance (ANOVA) followed by a Fisher’s least significant difference (LSD) multiple comparison test with no assumption of sphericity using the Geisser-Greenhouse correction to compare all mean nerve-to-background tissue ratios. The α value was set to 0.05 for all analyses. Results are presented as mean ± SD. All statistical analyses were performed using Prism (GraphPad). Full-width half-maximum calculations of fluorescence intensity line profiles were performed in Excel, where well-behaving (one peak with rising and falling edge) graphs allowed identification of half maximum values in both rising and falling edges and calculation of width therein.
SUPPLEMENTARY MATERIALS
stm.sciencemag.org/cgi/content/full/12/542/eaay0712/DC1
Materials and Methods
Fig. S1. Oxazine derivative library chemical structures.
Fig. S2. HPLC-MS and purity analysis of oxazine derivative library.
Fig. S3. 1D and 2D NMR characterization of LGW01-08.
Fig. S4. 1D and 2D NMR characterization of LGW05-75.
Fig. S5. 1D and 2D NMR characterization of LGW04-31.
Fig. S6. 1D and 2D NMR characterization of LGW03-76.
Fig. S7. Ex vivo nerve specificity screening of the oxazine library.
Fig. S8. In vivo direct administration nerve specificity screening of the oxazine library.
Fig. S9. In vivo systemic administration nerve specificity screening of the oxazine library.
Fig. S10. Potential lead NIR oxazine candidate structures and spectra.
Fig. S11. Biodistribution studies of potential lead NIR oxazine fluorophores.
Fig. S12. Imaging of nerves at depth after systemic administration of nerve-specific probes.
Fig. S13. Quantification of swine nerve imaging during minimally invasive surgery.
Table S1. HPLC-MS and purity analysis of oxazine derivative library.
Table S2. Oxazine derivative library optical properties.
Data file S1. Primary data.
Movie S1. Lumbar nerve visualization using LGW01-08 and da Vinci endoscope 2 hours after injection in swine.
Movie S2. Lumbar nerve visualization using LGW05-75 and da Vinci endoscope 5 hours after injection in swine.
Movie S3. Lumbar nerve visualization using LGW01-08 and da Vinci endoscope 20 min after injection in swine.
Movie S4. Buried lumbar nerve visualization using LGW01-08 and da Vinci endoscope 2 hours after injection in swine.
Movie S5. Buried lumbar nerve visualization using LGW05-75 and da Vinci endoscope 5 hours after injection in swine.
Movie S6. Iliac nerve visualization using LGW01-08 and da Vinci endoscope 15 min after injection in swine.
Movie S7. Iliac nerve visualization using LGW05-75 and da Vinci endoscope 5 hours after injection in swine.
Movie S8. Lumbar nerve sparing using LGW05-75 and da Vinci endoscope 6 hours after injection in swine.
This is an article distributed under the terms of the Science Journals Default License.
REFERENCES AND NOTES
- Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works
