Dextran-Mimetic Quantum Dots for Multimodal Macrophage Imaging In Vivo, Ex Vivo, and In Situ

Abstract

Macrophages are white blood cells with diverse functions contributing to a healthy immune response as well as the pathogenesis of cancer, osteoarthritis, atherosclerosis, and obesity. Due to their pleiotropic and dynamic nature, tools for imaging and tracking these cells at scales spanning the whole body down to microns could help to understand their role in disease states. Here we report fluorescent and radioisotopic quantum dots (QDs) for multimodal imaging of macrophage cells in vivo, ex vivo, and in situ. Macrophage specificity is imparted by click-conjugation to dextran, a biocompatible polysaccharide that natively targets these cell types. The emission spectral band of the crystalline semiconductor core was tuned to the near-infrared for optical imaging deep in tissue, and probes were covalently conjugated to radioactive iodine for nuclear imaging. The performance of these probes was compared with all-organic dextran probe analogues in terms of their capacity to target macrophages in visceral adipose tissue using in vivo positron emission tomography/computed tomography (PET/CT) imaging, in vivo fluorescence imaging, ex vivo fluorescence, post-mortem isotopic analyses, and optical microscopy. All probe classes exhibited equivalent physicochemical characteristics in aqueous solution and similar in vivo targeting specificity. However, dextran-mimetic QDs provided enhanced signal-to-noise ratio for improved optical quantification, long-term photostability, and resistance to chemical fixation. In addition, the vascular circulation time for the QD-based probes was extended 9-fold compared with dextran, likely due to differences in conformational flexibility. The enhanced photophysical and photochemical properties of dextran-mimetic QDs may accelerate applications in macrophage targeting, tracking, and imaging across broad resolution scales, particularly advancing capabilities in single-cell and single-molecule imaging and quantification.

Graphical Abstract

Immune cells, including T cells, natural killer cells, and macrophages, play a critical role in the immune response against pathogens and in immunoediting to prevent cancer.^1–3 Current efforts to develop imaging agents targeting these cells are driven by the need to understand their biological functions, to diagnose diseases related to the immune system, and to monitor the immune response to therapy.^4–6 Macrophage cells, in particular, are the subject of numerous ongoing investigations in the life sciences and in translational medicine due to their role in diverse physiological processes and diseases.^7–10 These myeloid leukocyte cells are classically known for their functions in innate and adaptive immunity, including the phagocytosis of pathogens and cell debris, display of antigens to T cells, and initiation of inflammatory processes. With extreme phenotypic plasticity, macrophages also promote repair in wounded and infected tissues^11–14 and can integrate and respond to signals across endocrine, inflammatory, immunological, and metabolic pathways in most mammalian tissues.^7,15,16 In chronic inflammatory disorders such as rheumatoid arthritis and type 2 diabetes, macrophages are implicated in initiating and maintaining inflammation.^8,17,18 In solid tumors, certain macrophage populations facilitate tumor cell invasion, metastasis, and angiogenesis and suppress adaptive immune responses.^7,19,20 For these reasons, concerted efforts are devoted to the development of targeted contrast agents and probes for macrophages for monitoring their prevalence, analyzing phenotypes, and understanding their underlying functions and dynamics.^21–24 One key missing technology at present is a probe which can resolve and quantify macrophages across broad resolution scales, from whole tissues and organisms to individual cells.

A variety of contrast agents accumulate in macrophages after administration to the body, including ferumoxides (e.g., Feridex), which are magnetic resonance contrast agents composed of superparamagnetic iron oxide nanoparticles coated with the polysaccharide dextran. After intravenous administration, these materials accumulate in both resident and infiltrating macrophages of the liver, spleen, adipose tissue, tumors, and atherosclerotic plaques.^21,24–29 A related agent is technetium-99m-labeled tilmanocept (Lymphoseek), a mannosylated dextran that also accumulates in phagocytotic cells including macrophages in lymph nodes after intratumoral injection.²⁷ Macrophage-targeted antibodies (anti-CD206, anti-CD11b) and small molecules, each with differential levels of macrophage selectivity, have also been conjugated to fluorophores, radioisotopes, and magnetic elements for imaging by fluorescence, positron emission tomography (PET) or single-photon emission computed tomography (SPECT), and magnetic resonance imaging (MRI), respectively.^30–34 Of these modalities, PET/SPECT is particularly valuable for sensitive, quantitative detection in vivo due to the molecular specificity of the associated contrast agents. For preclinical and ex vivo studies, fluorescence is a preferred modality due to its submicron resolution, low cost, and capacity for correlative evaluation with molecular stains through multispectral imaging and flow cytometry.^33,35–37

Dextran is a macrophage-targeting agent with numerous advantages compared with other such materials. As a biocompatible and water-soluble polysaccharide, it is used broadly across clinical medicine and the life sciences.^38,39 Dextran is composed of glucose units linked linearly through α-1,6 glycosidic bonds with both α-1,3 and α-1,4 cross-links.⁴⁰ As synthesized biogenically by several species of bacteria, the polymer has a broad molecular weight distribution which can be used to isolate a wide range of discrete sizes.⁴¹ Dextran is stable in aqueous solution, is not readily degraded by mammalian enzymes, and can be excreted intact from the body, eliciting little immunogenic, thrombotic, metabolic, or toxicological response across a wide range of organisms.⁴² Due to its biochemical inertness, dextran has been adopted as a standard tracer for fluid dynamics studies in blood and cells, for measuring fluid phase pinocytosis in cultured cells, and for simulating macromolecular crowding at high concentration.^43–45 Clinically, dextran has been used for more than 75 years as a plasma volume expander, with few side effects reported when intravenously administered at high doses.⁴⁶ Dextran is also a building block of drug formulations due to its high solubility and high density of functional groups for conjugation.⁴³ When administered to mammals at sizes larger than the renal filtration threshold (~40 kDa), its distribution is primarily determined by lectin receptor-mediated endocytosis through DC-SIGN and L-SIGN receptors, which are expressed on phagocytotic cells such as macrophages and dendritic cells.^47–49 Due to this targeting, together with the increased infiltration and activity of macrophages in inflamed tissues, dextran has been applied as a nanocarrier of imaging agents and therapies for macrophage cells in tumors, lymph nodes, adipose tissue, liver, spleen, and blood vessels.^21,47

We previously developed dextran-based macrophage probes with dual-modal fluorescence and PET contrast by conjugation to organic dyes and radiotracers.^50,51 These probes distributed to macrophages in inflamed adipose tissue of obese rodents but not in normal lean rodents. PET imaging yielded accurate localization of the probes within tissues, but fluorescence measurements did not quantitatively reflect the underlying biodistribution. This was evident for three different dyes, even with emission in the near-infrared (NIR) spectrum for which tissue is more permeable to light and exhibits lower autofluorescence background signals compared with the visible spectrum.^37,52,53 We attributed this limitation primarily to dye degradation in macrophages as well as the variable opacity of tissue, which interferes with optical measurements and comparisons between tissues.⁵⁰ An alternative contrast agent is a semiconductor quantum dot (QD) which is a colloidal crystalline fluorophore with orders of magnitude greater brightness, photostability, and chemical stability and with tunable emission throughout the visible and infrared spectra with high quantum yield.^54,55 These probes could alleviate problems with dyes and provide imaging and tracking capabilities at the single-molecule level.^53,56–59 However, as colloidal macromolecules, QDs are physically and biochemically dissimilar from dyes for derivatizing functional dextran conjugates for macrophage targeting.

In this work, we report a multimodal dextran-mimetic QD (Q-Dex, Figure 1a) with NIR emission at 800 nm wavelength as a macrophage contrast agent for multiscale, multimodal imaging through both in vivo fluorescence and PET and both ex vivo and in situ fluorescence microscopy. We coated QDs with a shell of dextran and compared their performance with a native dextran (Dex, Figure 1b) with matched hydrodynamic size, size distribution, and electrostatic charge. We labeled both materials with radioisotopes and labeled Dex with a fluorescent zwitterionic dye (ZW800) with matched emission wavelength as the QDs. We administered these agents to obese mice with macrophage-rich adipose tissues and measured nuclear and fluorescent signals using a variety of methods. PET/CT imaging and fluorescence/CT imaging reveal grossly similar distribution patterns, but pharmacokinetics measurements show significantly extended blood circulation times for Q-Dex. We attribute this difference to the greater rigidity of Q-Dex, corroborated by in vitro studies of diffusion in crowded media and by comparison with all-organic composite dextrans (C-Dex,Figure 1c) with core and shell of equivalent subunit dimensions to those of Q-Dex. Q-Dex showed improved quantification by macroscale fluorescence, long-duration resistance to photobleaching, and resistance to chemical fixatives that together enabled high-resolution subcellular 3D imaging in resected tissues. We expect that these probes can solve previous problems of organic dyes for high-resolution and quantitative optical imaging of macrophages.

Figure 1.

Chemical structures of multimodal macrophage-targeted dextran-mimetic QD probes (Q-Dex) and all-organic dextran probe analogues (Dex and C-Dex). (a) Structure of dextran-mimetic NIR QD probe composed of crystalline (core)shell (Hg_xCd_1–xSe)Cd_yZn_1–yS coated with a multidentate azide-functional polymer (P-IM-N₃) that is conjugated to dextran through a strain-promoted click reaction. Dextran is further labeled with radioactive iodine. (b) Native dextran probe conjugated to radiolabel and NIR dye ZW800. (c) Composite dextran probe with a (core)shell structure with equivalent subunit dimensions to those of Q-Dex, conjugated to radiolabel and ZW800.

RESULTS AND DISCUSSION

Dextran-Mimetic Quantum Dot Design and Synthesis.

We chose 800 nm as the target QD emission wavelength for high tissue permeation and sensitive detection by CCD and CMOS cameras in diverse optical imaging instruments. We synthesized conventional (core)shell (CdSe)-CdZnS nanocrystals but used cores doped with mercury to shift the emission from the visible to the infrared spectrum with high quantum yield.^56–58 Using 3.6 nm Hg_xCd_1–xSe cores, we generated (Hg_xCd_1–xSe)Cd_yZn_1–yS with quasi-spherical shape and crystalline diameter of 5.5 ± 0.9 nm by transmission electron microscopy (TEM, Figure 2a and b).^60,61 The QD emission band centered at 800 nm was near to those of dextran conjugates of ZW800 (Figure 2c). These NIR QDs were coated with multidentate polyacrylamido(histamine-co-triethylene glycol-co-azidotriethylene glycol) polymers (P-IM-N₃) developed previously to form stable aqueous (QD)P-IM-N₃ colloids with azide functionalization.⁶² Multidentate polymers have become effective ligands for QDs to generate homogeneously dispersed and compact aqueous QDs due to continual developments by researchers over the past decade.^63,64 A monolayer of 10 kDa dextran (Dex(l0)) functionalized with dibenzocyclooctyne (DBCO) was covalently deposited on the (QD)P-IM-N₃ surface by interfacial strain-promoted alkyne–azide cycloaddition (I-SPAAC).⁵⁹ The number of DBCO groups per Dex(10) was controlled to be near 1 to avoid cross-linking and minimize unreacted DBCO groups after conjugation. Dex(10) was also conjugated to the Bolton-Hunter HPP reagent for radiolabeling with ¹²⁴I or ¹²⁵I. C-Dex was prepared using the same Dex(10)-DBCO-HPP “shell” used for Q-Dex, but with a 70 kDa “core” azide-functional dextran (Dex(70)-N₃) in the place of (QD)P-IM-N₃. Native Dex was purified from 250 kDa dextran (Dex(250)). Dex and C-Dex were both conjugated to ZW800 and also labeled with ¹²⁴I or ¹²⁵I. Each dextran building block was purified by gel permeation chromatography (GPC) from heterogeneous commercial products prior to allylfunctionalization and amination by cysteamine-ene click reactions for labeling using N-hydroxysuccinimidyl (NHS) ester reagents (Supporting Schemes S1, S2, and S3), with chemical structures confirmed by ¹H nuclear magnetic resonance (¹H NMR) (Supporting Figure S1).

Figure 2.

Characterization of Dex, C-Dex, and Q-Dex. (a) TEM micrographs of Hg_xCd_1–xSe QD cores (upper) and (core)shell (Hg_xCd_1–xSe)Cd_yZn_1–yS (lower) with 800 nm emission (QD800). (b) TEM size distributions with Gaussian fit (red line), showing mean crystalline diameter (D_C). (c) Fluorescence emission spectra of Dex, C-Dex, and Q-Dex. (d) GPC chromatograms of Dex, C-Dex, Q-Dex, and the building blocks Dex(70) and (QD)P-IM-N₃. Elution times of protein standards and gold colloid are indicated by hydrodynamic diameters on the top x-axis. Representative FCS autocorrelation functions are shown for the three materials dispersed in (e) PBS or (f) PBS with 30 wt % Ficoll, with calculated diffusion coefficients shown in parts g and h, respectively. Statistical significance was determined by Student’s t test: n.s. is not significant, * P < 0.05, ** P < 0.01.

Physicochemical Properties.

Q-Dex, Dex, and C-Dex were designed in tandem for equivalent biochemistry, size distribution, and charge, with the expectation that these characteristics together dictate performance as in vivo macrophage probes (Figure 1).^65,66 Sizes were designed to match that of Dex(250), a dextran size range that was previously found to target visceral adipose tissue macrophages after intraperitoneal administration.⁵¹ Q-Dex, Dex, and C-Dex all measured near 24 nm hydrodynamic diameter (D_H) in phosphate buffered saline (PBS, pH 7.4) by GPC (Figure 2d, Supporting Table S1) calibrated by globular proteins and gold nanoparticles (Supporting Figure S2). For Q-Dex, this total size derived from a 13 nm (QD)P-IM-N₃ core with a saturating Dex(10) shell. As expected, D_H was similar for a C-Dex composite composed of a GPC-purified 13 nm Dex(70) core with a saturating Dex(10) shell. We note that GPC purification of the dextran building blocks was necessary to reduce the substantial polydispersity of the raw commercial products with ~80% relative standard deviation (RSD) in D_H,⁵⁰ resulting in a total 13–17% D_H RSD. Like GPC, fluorescence correlation spectroscopy (FCS) and dynamic light scattering (DLS) showed similar sizes in PBS (Supporting Table S1, Supporting Figure S3). Zeta potentiometry measurements also showed similar electrostatic charges for the three materials (−3 to −4 mV, Supporting Table S1) in sodium phosphate buffer (pH 7.4, 10 mM).

While the physicochemical properties of the three dextran-based materials were nearly identical in aqueous solutions, differences were measured in conditions of high macromolecular crowding simulating hydrogel-like tissue microenvironments. FCS autocorrelation functions were fit using a Brownian diffusion model for the three materials either in PBS alone (Figure 2e) or in PBS with the crowding agent Ficoll 70 (30 wt % Figure 2f).⁴⁵ While calculated diffusion coefficients were statistically indistinguishable in PBS (Figure 2g), macromolecular crowding reduced the diffusion coefficients of Q-Dex and C-Dex to levels nearly 7-fold lower than that of Dex (Figure 2h). This is surprising based on the structural and biochemical similarities between the materials, as 88% of the volume of Q-Dex is dextran, and 99% of the volume derives from flexible organic polymers (only ~1% is the rigid QD crystal itself). However, diffusion differences have been observed previously for different classes of macromolecules with equivalent physical dimensions.⁶⁷ The similarity between Q-Dex and C-Dex shows that the result likely derives from the underlying material conformational flexibility, as the degree of cross-links from branches (~5% of subunits) in native Dex should be much lower than that of the two composites.⁴⁰ Thermal and convective transport of colloids through crowded and porous matrices can be enhanced for flexible materials, which can manifest in dissimilarities in pharmacokinetics and biodistribution.^68–70

Pharmacokinetics and Biodistribution.

We next administered the three multimodal probes to obese C57BL/6 mice intravenously (IV, tail vein) to determine blood circulation times assessed by blood radioactivity after cheek bleeds (Figure 3a and b). Radioactivity quantification in blood was determined via a standard calibration curve (Supporting Figure S4). The monoexponential elimination half-life of Dex was 38 min, which was much shorter than that of the two composite dextrans. The blood half-life increased 7.6-fold for C-Dex (4.74 h; P = 0.0008) and 8.9-fold for Q-Dex (5.58 h; P = 0.0002). These differences were unexpected based on their similar sizes and charges in buffered saline (Figure 2d and g) but correlate with mobility differences in crowded media (Figure 2h). Because crowded media more closely reflect the meshlike glycocalyx-lining of the capillary endothelium, it is possible that the higher flexibility of native Dex allows its permeation through vascular pores of reticuloendothelial system organs such as the liver and spleen, where the majority of exogeneous intravenous nanomaterials distribute.⁶⁹ We note that the blood half-life difference between Q-Dex and C-Dex was insignificant (P = 0.29), similar to their equivalent diffusion coefficients in crowded media (Figure 2h).

Figure 3.

Pharmacokinetics and biodistribution of Dex, C-Dex, and Q-Dex by gamma well counting (GWC) of harvested tissues. (a) Biodistribution of probe (green) in reticuloendothelial system organs (liver and spleen) after intravenous (IV) injection in obese C57BL/6 mice. (b) Concentrations of the three probes in blood after IV administration measured by cheek bleeds at 40 min, 1.5 h, 4 h, 7 h, and 24 h postinjection. Elimination half-life is shown as t_1/2. Data are presented as mean ± SD for n = 3 or 4 independent experiments. (c) Tissue biodistribution 24 h after IV administration measured by GWC. Tissues shown are blood, visceral adipose tissue (VAT), kidney, lung, heart, liver, and spleen. Liver and spleen are plotted separately due to large differences in absolute concentrations. Data are presented as mean ± SE for n = 4. (d) Biodistribution of probe in VAT, liver, and spleen after IP injection. (e) Tissue biodistributions of the three probes after intraperitoneal administration measured by GWC. Data are presented as mean ± SE for n = 9. (f) Cell distribution via fluorescence-based flow cytometry, showing the fraction of CD11b-positive VAT cells that are positive for the probes. (g) Flow cytometry fraction of probe-positive VAT cells that are CD11b-positive. Flow cytometry data are presented as mean ± SE for n = 4. * P < 0.05, ** P < 0.01, *** P < 0.001, **** P < 0.0001.

Twenty-four hours after administration, at which point more than 95% of each material was eliminated from circulation, the animals were sacrificed and tissues were collected for analysis by calibrated gamma well counting (GWC, Figure 3c). Dex was primarily located in the liver and spleen, where concentrations were more than 60-fold greater than any other tissue, consistent with previous studies.^51,72 Compared with Dex, both C-Dex and Q-Dex exhibited higher uptake in other organs, including kidney, lung, and heart, but the absolute quantity of material in these organs remained small compared with liver and spleen. For all materials, little uptake was measured in visceral adipose tissue (VAT), which has low blood perfusion compared with the other tissues. While differences in uptake by the liver were not significant across the three different materials (P > 0.071), uptake in the spleen was much higher for C-Dex and Q-Dex compared with Dex, with C-Dex exceeding Dex by nearly 8-fold. The spleen has previously been observed to collect longer circulating nanomaterials,⁷² as it receives less total cardiac output (~6%) than the liver (~25%) but is rich in phagocytotic cells that are directly perfused by blood.^73,74 This perfusion through a crowded matrix may result in probe-specific entrapment like that observed by static FCS (Figure 2h). The spleen also collects senescent and deformed red blood cells (RBCs), which can facilitate splenic accumulation of circulating matter that adheres to RBCs.^75–77 We compared the binding of Dex, C-Dex, and Q-Dex probes to RBCs in vitro relative to polystyrene beads as a positive control and IgG as a negative control (Supporting Figure S5). Binding was minimal at 15 min, but by 60 min of incubation, all three probes were found to bind to some degree to RBCs, with C-Dex showing significantly higher RBC adsorption than Dex and Q-Dex. This difference may explain the unusually high spleen uptake of C-Dex. We note that mouse and human spleens differ substantially in anatomy, so it is uncertain whether the observed distribution differences in the spleen would occur in a clinical scenario.⁷⁸

The biodistribution of nanomaterials can dramatically differ depending on the administration route.⁶⁸ We previously observed that for molecular weights greater than ~20 kDa, dextran distributes primarily to macrophages in VAT after intraperitoneal (IP) injection, with some distribution to liver and spleen, as depicted in Figure 3d. This effect occurs due to direct perfusion of interstitial spaces of peritoneal tissues such as VAT,⁷⁹ which are enriched with inflammatory macrophages in the obese state. VAT macrophage targeting was not observed in the lean state in which macrophages are less prevalent in VAT.^50,51 By time-course in vivo dynamic PET imaging at 2–24 h after IP administration (Supporting Figure S6), we found that distribution to VAT and liver occurred rapidly, with little change after 2 h, consistent with efficient dextran uptake in VAT macrophages.^21,50 We previously evaluated the targeting capacity of Dex in obese mice after IP administration and found that it was macrophage-selective, with VAT uptake eliminated when using Dex controls, including when Dex was conjugated to polyethylene glycol to block affinity for glucose groups by lectins and when Dex was low in molecular weight (10 kDa), which abrogates avidity toward receptors.⁵¹

Figure 3e shows quantitative biodistributions for Dex, C-Dex, and Q-Dex probes measured by GWC of resected tissue 2 h after IP administration. Compared with IV injection, concentrations in VAT were higher by factors of 15-fold for Dex, 14-fold for C-Dex, and 20-fold for Q-Dex. Correspondingly, concentrations in liver were lower for IP injection compared with IV injection, by factors of 1.6-fold for Dex, 1.9-fold for C-Dex, and 2.7-fold for Q-Dex. Interestingly, trends in organ specificity for the three probes mirrored those observed after IV injection, with Dex having the highest liver uptake, C-Dex having the highest spleen uptake, and all three showing statistically equivalent VAT uptake. Liver and spleen uptake likely resulted from some fraction of the administered peritoneal dose entering the bloodstream. Notably, Q-Dex exhibited the lowest liver uptake of the three materials. The tissues shown in Figure 3 were previously found to be the primary targets for Dex;^50,51 however, we also used PET/CT imaging to confirm that noninflamed tissues (e.g., skin and muscle) were not targeted (Supporting Figure S7 and see images below). The total recovered radioactivity of all resected tissues for Dex, C-Dex, and Q-Dex probes was less than 50% for all probes at 24 h after injection, likely due to hepatobiliary excretion (Supporting Figure S8).

We further note that, consistent with previous studies, most VAT cells accumulating these probes expressed CD11b, a marker of myeloid populations.^51,80,81 There were no statistically significant differences across the three probes for the fraction of CD11b+ cells labeled by probes (Figure 3f) or the fraction of probe-labeled cells expressing CD11b (Figure 3g), which was not surprising based on the equivalent biochemical functionalization of the probes. This suggests that macrophage selectivity is equivalent for the three materials so biodistribution differences likely arise due to differences in transport alone. It should be noted that the selectivity of Dex derivatives toward macrophages and their subpopulations relative to other cell types has not been comprehensively evaluated, and our knowledge of the repertoire of receptors responsible for cell-type tropism is incomplete. One challenge is the diversity of administration routes, dosages, and disease indications for which dextran is applied. A second challenge is the difficulty of precisely classifying macrophage subpopulations which express a spectrum of biomarker profiles with significant overlap with other leukocytes, such as monocytes and dendritic cells. Therefore, for one of the materials (Dex), we evaluated the cell distribution in depth in two VAT depots (left gonadal and left perirenal) from obese mice after IP injection in the left side. Cells in VAT beds were stained for monocyte/macrophage lineage markers (CD11b, F4/80, Ly6C) as well as non-macrophage markers (CD3, CD19, Ter119, NK1.1, Ly6G) for comparison (Supporting Figures S9 and S10). In both tissue beds, more than 60% of Dex-positive cells were inflammatory, infiltrating monocyte/macrophage populations (Ly6C+). These Ly6C+ cells have been identified as recently recruited macrophages that are uniformly distributed throughout VAT in the obese state, but not in the lean state, and are a primary cell type of interest for cytological monitoring and therapeutic targeting of adipose changes in obesity.⁸² Preferential uptake of Dex by macrophages relative to adipocyte-like cells was likewise observed in cultured cells, although no selectivity was observed for macrophage cells based on inflammatory phenotype (Supporting Figure S11).

Ex Vivo Whole-Tissue Fluorescence Imaging and Quantification.

Wide field fluorescence imaging of live resected tissues (PerkinElmer IVIS with 745 nm excitation) was used to analyze biodistribution patterns after IP administration. Fluorescence measurements coarsely correlated with gold standard GWC for the three probes (Supporting Figure S12); however, large disparities were observed in absolute values, especially when comparing across different organs, such as between VAT and spleen. These disparities are likely due to limitations of fluorescence in terms of depth penetration, background autofluorescence, detection range, sensitivity, and photostability.⁸³ Figure 4a and b shows Bland–Altman plots for evaluating agreement between fluorescence measurements and GWC measurements across all target VAT issues for Dex and Q-Dex probes. Q-Dex showed 2.3-fold higher overall agreement based on the 95% agreement limit (0.20) compared with Dex (0.46). Q-Dex also showed significantly less systematic bias based on median offsets (−0.014) than Dex (−0.11), which underestimated the probe quantity in most tissues. C-Dex showed similar metrics as Dex (Supporting Figure S13). In contrast with white-colored VAT, Q-Dex performance improvements compared with Dex were less dramatic in red-colored tissues such as liver and spleen, which may reflect a larger impact of tissue autofluorescence, light absorption, and light penetration in these tissues (Supporting Figure S14). We also previously reported a better correspondence overall for white tissues compared with red tissues for three classes of dye labels.⁵⁰

Figure 4.

Biodistribution quantification accuracy by fluorescence imaging. (a and b) Bland–Altman plots show normalized fluorescence (FL)-based biodistribution for Dex and Q-Dex relative to GWC in VAT. Data points represent tissues following IP administration and excision. Orange lines are median values, dashed red lines are 95% confidence intervals, and blue lines are linear regressions. (c) Photobleaching of Dex and Q-Dex in living resected VAT or liver measured by repeated fluorescence imaging with radiant efficiency normalized to the first measurement. Mean values are shown as data points with standard deviations as shaded areas.

Results in living tissues were consistent with measurements in tissue phantoms. QDs could be detected over a deconvolved autofluorescence background at lower concentration compared with the ZW800 dye both in PBS and in homogenates of VAT and liver (Supporting Figure S15). In homogenates, signals were linear across a broader range of concentrations for QDs compared with ZW800 (Supporting Figure S16). The dye exhibited poorer linearity and limit of detection (LOD) in tissue homogenates compared with PBS, whereas both parameters were slightly better for QDs in homogenates than in PBS. We attribute this difference to the biochemical lability of dyes relative to QDs deriving from the polymer encapsulation of QDs that prevents impacts from surrounding solutes.

The higher overall accuracy of fluorescence quantification for Q-Dex compared with C-Dex and Dex likely derives from the photophysical differences between QDs and dyes. While the emission spectra were generally similar (Figure 1c), Q-Dex had a 6- to 7-fold higher brightness per particle when excited at 745 nm due to higher quantum yield (47%) compared with Dex (6.1%) and C-Dex (4.8%), even with conjugation of ~9 dyes per dextran (Supporting Table S2).^52,54 Differences were also observed in photostability, which can substantially impact fluorescence quantification, especially for organic dyes which can rapidly degrade during continuous photoexcitation.^50,52,53 No significant photobleaching was detected for Q-Dex after 98 sequential image collections in live VAT, and there was even a significant increase of emission intensity in liver likely deriving from photoenhancement (Figure 4c).^84,85 In contrast, there was a progressive radiant efficiency loss for Dex probes with each image acquired in both VAT and liver.

Multiscale Multimodal Imaging.

Figure 5 displays the diverse array of nuclear and fluorescence-based imaging modalities enabled by macrophage-targeted Dex and Q-Dex probes after IP administration of 1.0 nmol of each probe to obese mice. Figure 5a shows 3D tomographic reconstructions of PET images overlaid on CT images of the whole peritoneal space of living mice under isoflurane anesthesia in a supine position. The localization patterns for Q-Dex and Dex were grossly dissimilar, with primary distribution of Dex to liver in the anterior cavity and that of Q-Dex to large adipose depots in the peripheral regions of the cavity. The same mice were imaged by 3D tomographic NIR fluorescence with correlative CT, using identical excitation bands, emission filters, and imaging conditions. The reconstructed images are shown in Figure 5b, showing signal throughout the body cavity. The majority of the Q-Dex signal was localized peripherally compared with Dex, which is consistent with PET images. Figure 5c shows that once the skin covering the body cavity was removed, fluorescence of Q-Dex was distributed throughout VAT within the body cavity. This was not the case for Dex, which showed only a single bright region that did not spatially colocalize with the PET pattern. Figure 5d shows PET and FL images of resected organs of interest; contrast correlated grossly between the two modalities, but with higher agreement for the Q-Dex probes compared with Dex probes (Figure 4a and b).

Figure 5.

2D and 3D nuclear and optical imaging via PET and fluorescence (FL) imaging of macrophage-targeted probes in obese mice in vivo, ex vivo, and in situ. (a) Reconstructed 3D images of multimodal Dex or Q-Dex probes in obese mice acquired with micro-PET/CT imaging approximately 24 h after intraperitoneal injection (1.0 nmol). Cross-sectional slices of the corresponding 3D PET/CT images from z-positions centered within the subjects. (b) Deconvolved 3D FL images overlaid on CT images for multimodal probe Dex or Q-Dex, acquired in vivo (IVIS) for the same mice as in panel a. Imaging regions of interest are centered on the peritoneal cavity. Cross-sectional slices of the corresponding 3D FL/CT images are from z-positions centered in the subjects. Note that the PET and FL images are not precisely registered due to moving the mice between two instruments. (c) 2D FL images of mice live or after sacrifice and surgical exposure of the peritoneal cavity to reveal peritoneal tissues. (d) FL images and PET coronal slice images of resected VAT and liver ex vivo. (e) Bright field (BF) and wide field FL microscopy images of excised, fixed VAT with 750 nm excitation and 785/40 emission filters. Only the Q-Dex probe signal is measurable over background. Zoomed in region shows punctate patterns consistent with intracellular vesicles. Scale bars: 30 μm.

For multimodal imaging of radioisotopic and fluorescent probes, correlative tissue imaging by microscopy typically necessitates chemical fixation to preserve tissue ultrastructure to provide sufficient time for radioactive decay. We fixed VAT that was excised from obese mice administered the Dex-based probes and imaged it intact without physical sectioning via fluorescence wide field microscopy at 100× magnification (Figure 5e). Dye fluorescence was generally undetectable throughout VAT for Dex, and images showed similar patterns as tissues in which no probes were administered. In a small number of areas with detectable signals, intensity disappeared too rapidly to image due to photobleaching, which is commonly observed for NIR organic dyes.^50–52 In contrast, VAT from mice administered Q-Dex exhibited bright, stable signals with intensity much higher than background, localized in small interstitial cells consistent with macrophages and excluded from larger adipocytes. This localization pattern was consistent with that previously identified using fluorescent Dex conjugates in freshly sliced specimens without fixation.⁸⁶ It is important that these fixed tissues could be easily imaged intact without slicing, which extensively deforms low-density adipose tissue and allows contrast agents to diffuse and spatially redistribute. The punctate appearances and micron-scale sizes of Q-Dex-labeled structures were consistent with endosomal and lysosomal compartments in macrophages, which was consistent with lysosomal tracker dye colocalization measured in cultured inflammatory macrophages (Pearson’s r = 0.833, Supporting Figure S17). We validated this through high-resolution imaging of freshly resected, intact live tissue. As shown in Figure 6, 3D spinning disk confocal imaging revealed individual cells with similar fluorescence localization patterns to those from fixed specimens, confirming the contrast specificity.

Figure 6.

Spinning disk confocal microscopic imaging of macrophages in live adipose tissue ex vivo. Obese mice were administered Q-Dex (nonradioactive) IP before VAT was resected and stained live with Hoechst prior to fluorescence imaging. Hoechst is shown in blue, and Q-Dex is red. Scale bars: 30 μm.

Cytotoxicity.

To investigate if these probes might induce toxic effects after administration, cell lines relevant to the uptake pattern after IP administration were assayed for viability after treatment for 24 h with Dex, C-Dex, Q-Dex, and QD(P-IM-N₃). Cells included human and mouse macrophages as well as mouse 3T3-L1 cells differentiated to adipocytes. There was no significant loss of cell viability measured for the highest probe concentrations up to 50 nM (Supporting Figure S18), which exceeded the maximum tissue concentration measured after IP administration (Figure 3e). It was expected that toxic effects due to dextran would be minimal;⁸⁷ however, Q-Dex contains heavy metal atoms (cadmium and a small amount of mercury) that could be toxic if they were to leach from the composite.⁸⁸ Cells indeed showed significant viability loss when exposed to Cd²⁺ ion concentrations of 10 μM and above. These results suggest that metal atom dissociation from Q-Dex is unlikely to occur over a 24 h time scale, which is consistent with the protective effects of the adsorbed organic polymer and dextran shell.⁸⁹

DISCUSSION

We described the engineering of dextran-mimetic QDs with NIR emission and nuclear contrast for multiscale in vivo, ex vivo, and in situ imaging of macrophages in mice. In comparison with native 250 kDa dextran, these probes have equivalent hydrodynamic size near 24 nm (Figure 2) and exhibit equivalent VAT and macrophage targeting following regional IP administration (Figure 3). Equivalence in physicochemical and biochemical properties is expected to yield equivalent biophysical transport and receptor-dependent targeting.^46,49 The retention of the biochemical selectivity of dextran when applied as a shell is consistent with previous findings of lectin-induced aggregation and macrophage uptake of dextran–nanoparticle conjugates.^21,32

A surprising finding was that the circulation time for C-Dex and Q-Dex was extended 7.6–8.9-fold compared with Dex (Figure 3b). Because RBCs bound to C-Dex to a significantly greater extent than Dex (Supporting Figure S5), it could be inferred that C-Dex adsorption to these long-circulating cells extended its circulation time in blood.^75–77 Indeed, this can explain the preferential distribution of C-Dex to the RBC-rich spleen (Figure 3c and e); however, this effect alone is insufficient to explain the extended circulation time of Q-Dex, which showed equivalent RBC binding as Dex. Because C-Dex was similar to Q-Dex in physical construction and surface biochemistry, it is likely that extended circulation times derive in part from internal material structure. Native Dex is a modestly branched polymer and is more deformable than globular proteins and synthetic polysaccharides such as Ficoll.⁹⁰ For C-Dex and Q-Dex, surface Dex(10) polymers should be conformationally constrained compared with native Dex due to added covalent cross-links and circumferential deposition as an ordered (core)shell structure with reduced capacity to deform. Evidence for this difference was observed by FCS in a crowded solution of 30 wt % Ficoll (Figure 2), a medium composed of 20% excluded volume of polydisperse physical obstructions.^91,92 In this environment, 24 nm probes are spatially confined but flexible subdomains can diffuse through constricted channels. Dextran is known to deform and expands in solution to a much greater degree than other macromolecules with equivalent molecular weight (MW): the hydrodynamic size of dextran (D_H ∝ MW^0.437) is larger than the size of more cross-linked polysaccharides such as Ficoll (D_H ∝ MW^0.427) and globular proteins (D_H ∝ MW^0.386).⁹⁰ As a result, dextran permeates more through biological pores compared with more rigid macromolecules with equivalent D_H. In the prototypical example of glomerular capillary wall permselectivity, dextran has 11-fold higher urinary clearance than globular proteins and 2-fold higher clearance than Ficoll for equivalent D_H near 6 nm. More generalized effects of macromolecular rigidity on biological fate and transport have been reviewed in detail in the context of drug delivery vectors.^69,70

For macromolecules with D_H > 10 nm such as Dex, C-Dex, and Q-Dex, liver clearance is the main route of elimination determining circulation time.⁶⁵ Rigidity differences for these ~24 nm probes should result in transport differences through crowded pores of the hepatic capillary endothelium, which has been shown to exhibit size-dependent filtration of dextrans.⁹³ However, compared with renal filtration rates, less is known about how physicochemical parameters such as rigidity dictate liver uptake rate. While the circulation times of more rigid Ficoll (~30 min for 70 kDa)⁹⁴ have been reported to be shorter than those of dextran with equal D_H (~50–100 min for 40–70 kDa),⁹⁵ direct comparisons have not been reported and absolute t_1/2 varies widely across reports, with substantial dose dependence.⁹⁶ In fact, circulation times have been reported to span minutes to nearly a day for dextran-coated nanoparticles, even for materials with similar sizes and doses.⁹⁷ For proteins with sizes exceeding the renal filtration threshold, elimination rates also span orders of magnitude, reflecting biochemically specific elimination processes. For these reasons, comparative studies using materials with equivalent biochemical properties and size are needed. This report describes direct comparisons between native dextran and dextran as a colloidal surface coating.^95,96 In another notable direct comparison, the Muzykantov group used multiple nanoparticles with 150 and 300 nm sizes coated with equivalent antibodies and observed that more flexible nanoparticles were able to access spatially confined ~50 nm endothelial subdomains (caveolae) in lungs that were inaccessible to more rigid nanoparticles.⁹⁸ However, they observed an opposite trend in clearance rate as our materials, with more rigid materials clearing faster than those with more flexible structures. Clearly, flexibility is a fundamental nanomaterial design parameter but may be interdependent on size and biochemistry. Additional quantitative studies are needed to understand mechanical effects in biological microenvironments and interfaces for which receptor-mediated, active, and facilitated transport processes occur that cannot be simulated in vitro. For Q-Dex, the enhanced circulation time for these more rigid composites may benefit applications of dextrans as size-specific permselectivity probes and vascular tracers in animal models.^27,46–48

Dextran-mimetic Q-Dex exhibited improved NIR fluorescence quantification compared with dye-labeled Dex probes (Figure 4). The use of fluorescence imaging to evaluate nanomaterial biodistribution and pharmacokinetics has become increasingly prevalent although results can be questionable due to poor quantitative capabilities intrinsic to the techniques.^71,82 As we showed previously,⁵⁰ fluorescence-based biodistribution across different tissues was highly inaccurate, especially when evaluating red-colored tissues such as liver for which opacity obstructs a large fraction of emitted fluorescence. This is particularly limiting when using dyes that may be susceptible to photophysical and chemical modifications in biological compartments such as lysosomes. Direct comparisons with dyes using equivalent excitation and emission conditions showed that QDs had improved photostability (Figure 4c) and better detection limits and linear range (Supporting Figure S16), resulting in more accurate quantification based on fluorescence (Figure 4a and b). We attribute the better performance of QDs to their higher brightness (Supporting Table S2) and inertness to biological microenvironments. Unlike dyes, QDs are protected chemically from the acidic, oxidative, and enzymatic milieu of lysosomal compartments due to their multilayer crystalline and polymeric shell, and have been found to retain fluorescence years after in vivo administration.⁹⁹ Dyes, as we showed previously, are susceptible to environmental changes, as well as metabolic processing, photochemical damage, and self-quenching, effects that are difficult to predict and simulate in vitro.^50,100 Biodistribution quantification can also be performed by tissue homogenization and extraction to solubilize tracers and probes in solutions compatible with optical or mass spectroscopies. These methods are well-established for metal-based labels, but fluorophore quantification can still be obscured by metabolic and photochemical processes that occur prior to extraction. Furthermore, homogenization is irreversible and results in loss of potentially valuable localization information.

A wide range of NIR QD materials classes has been reported, including CdTe/CdS, PbS/CdS, InAs, and CuInS.^101–104 The HgCdSe-based QDs that we developed are advantageous for their high quantum yield in water (near 50%) and compact sizes (~5–6 nm), a combination not provided by other material classes.¹⁰⁴ The small sizes are important to allow a high volume density of targeting materials in the final composite. Most importantly the emission can be continuously tuned without changing core size,⁶¹ even to the short-wave infrared spectrum near 1400 nm where scattering and autofluorescence are even more reduced compared with the NIR,¹⁰⁴ although less common photodetectors and optics are needed for imaging.^53,105 We recently showed that shifting the emission to the SWIR resulted in a 10–55-fold improvement in signal-to-background signals in a variety of cell and tissue imaging applications.¹⁰⁴ Ongoing efforts to adopt these newer imaging systems, together with ones based on photoacoustic techniques,¹⁰⁶ provide opportunities to bridge the macroscale to microscale using optical measurements without the need for radioisotopes, which pose experimental and safety challenges.

We believe that Q-Dex probes will particularly benefit macrophage analysis in preclinical animal models for the study of the multitude of diseases impacted by macrophage cells. QDs have been successful as cell tracers due to their long-term stability and capacity for cell labeling and multispectral encoding.^107,108 A current need is to noninvasively analyze macrophage presence and phenotype in living tissue without end-point sacrifice for analysis by flow cytometry.^24,28 Further, more detailed knowledge of microscopic macrophage microenvironments such as adipose tissue crown-like structures,¹⁰⁹ revealed through high-resolution imaging, would be particularly beneficial to understand pathophysiological processes within tissue. The high sensitivity of PET together with single-cell imaging capabilities of QDs could make these multiscale correlations possible in the same tissues. PET/fluorescence probes have major advantages over high-resolution modalities such as magnetic resonance imaging, which requires orders of magnitude higher doses of contrast agents, which can lead to toxic effects of the probes.¹¹⁰ We note that while these probes are composed of cores containing heavy metal elements which provide exceptional optical properties, we did not observe cytotoxicity over 24 h of exposure at relevant concentrations needed for high-contrast imaging (Supporting Figure S18). Nevertheless, for long-duration studies in animals and eventual translational studies, compositions should be switched to more biologically friendly materials such as CuInS₂,¹¹¹ which would not lead to potential release of toxic elements over long durations in vivo.

CONCLUSIONS

In summary, we reported a NIR QD-based multimodal probe (Q-Dex) for efficient targeting and multiscale imaging of macrophages in obese rodents in vivo, ex vivo, and in situ. The Q-Dex probe showed similar targeting at the level of cells and tissues in comparison with widely used native dextran; however, its blood half-life was extended nearly 9-fold, likely due to reduced mobility in crowded microenvironments. Biodistribution quantification via optical imaging was significantly more accurate for Q-Dex, especially in white tissues, which further enabled high-resolution tissue imaging of macrophage cells. These probes set a precedent for the development of in vivo analytical tools for macrophages. Our ongoing efforts aim to further develop these probes for noninvasive quantification and localization of specific macrophage subpopulations that are predictive of disease progression or therapy response. Additional studies are needed to evaluate the pharmacokinetics of elimination, in vivo acute and chronic toxicity profiles, and macrophage-subtype selectivity dependence on probe structure and biochemistry. We expect that our designs can be generalized for the development of macrophage-targeting agents based not only on QDs but also on translational nanomaterials composed of carbon and silicon, as well as polymers.

METHODS

Synthesis of Dextran-Amine.

Amine-functionalized dextrans were prepared in a two-step process starting with commercial dextrans with molecular weights of 10 kDa, 70 kDa, and 250 kDa. First, dextran was reacted with allyl bromide with base catalyst to generate dextran-allyl. Then, dextran-amines were generated by thiol–ene reactions. Detailed synthesis and characterization methods, in addition to important considerations for the safe handling of radioactive materials, are provided in the Supporting Information.

Functionalization of Dextran-Amine.

Dextran-amines were conjugated to DBCO, ZW800, and HPP functional groups via NHS ester chemistry in anhydrous dimethyl sulfoxide overnight with base catalyst. Detailed synthesis, purification, and characterization methods are provided in the Supporting Information.

QD Synthesis and Polymer Coating.

NIR QDs were synthesized from CdSe QD cores with partial metal exchange of cadmium to mercury before layer-by-layer growth of CdS and ZnS shells. The resulting QD800 was further coated with the multidentate polymer P-IM-N₃ to disperse the QDs in aqueous solution and provide azide functional groups. Synthesis, purification, and characterization methods are described in the Supporting Information.

Q-Dex and C-Dex Syntheses.

Azide-functionalized (QD)P-IM-N₃ or Dex(70)-N₃ was dissolved in PBS and mixed with an excess of Dex(10) functionalized with DBCO and HPP. The products were purified by centrifugal filtration or gel permeation chromatography. Detailed synthesis, purification, and characterization methods are provided in the Supporting Information.

Determination of Dye Number and Quantum Yield.

The number of organic dyes per dextran was calculated from absorption spectra of solutions of known dextran concentration in PBS, applying the Beer–Lambert law to determine the concentration of dyes. Fluorescence quantum yields (Φ) were calculated from dilute solutions of each probe using the following formula:⁵⁷

$${\text{Φ}}_{\text{sample}}={\text{Φ}}_{\text{standard}}·\frac{F_{\text{sample}}}{F_{\text{standard}}}·\frac{A_{\text{standard}}}{A_{\text{sample}}}·{\left(\frac{n_{\text{sample}}}{n_{\text{standard}}}\right)}^2$$

where F is the integrated area of the fluorescence emission spectrum, A is the absorbance at the excitation wavelength, and n is the refractive index of the solvent. Free ZW800 in PBS was excited at 768 nm and was used as the standard using parameter values provided by the manufacturer (Φ = 13.5%).

Animals and Diets.

Male C57BL/6 mice were purchased from Jackson Laboratory and given a high-fat diet (HFD, D12492 [Research Diets, Inc.], 60% kcal fat) ad libitum for 8 weeks to reach obesity status prior to use. All animals were maintained in temperature- and humidity-controlled housing, with a 12 h–12 h light–dark switching cycle. All animal procedures were approved by the Institutional Animal Care and Use Committee at the University of Illinois.

Intravenous Administration.

Obese mice in groups of four were injected IV with one of the three probes labeled with radioactive iodine in 100 μL of saline. Minimum blood samples (~20 μL) were collected via cheek bleeding at 40 min and 1.5, 4, and 7 h postinjection. Mice were sacrificed at 24 h through cervical dislocation, and tissues including blood were collected for biodistribution analysis. All blood samples were analyzed via GWC and quantified as % ID g⁻¹. Blood data were fit to a single exponential decay to estimate the circulation half-life.

Intraperitoneal Administration, In Vivo Imaging, and Ex Vivo Imaging.

Twenty-seven obese mice were divided randomly into three groups and administered either Dex, C-Dex, or Q-Dex via IP injection in 100 μL of PBS on the left side. At specific times after injection, mice were anesthetized by 2% isoflurane inhalation and imaged via PET/CT using a hybrid microPET-SPECT-CT small animal scanner (inveon, Siemens Healthcare). For dynamic studies, mice were imaged at 2, 4, 8, and 24 h. For single-time point imaging, mice were imaged 2 h after injection. Imaging consisted of a 15 min static PET with 20% energy window centered at 511 keV and a high-resolution anatomic CT. After PET/CT imaging, animals were subject to fluorescence 3D imaging using an IVIS Spectrum-CT (PerkinElmer) with 745 nm excitation followed by 2D imaging with the same excitation. Regions of interest (ROIs), including both left and right abdomen and liver, were selected for fluorescence 3D imaging. The mice were then sacrificed by cervical dislocation, and the peritoneum was opened for IVIS 2D imaging. Tissues of interest were excised and imaged again in 2D using the same protocol used for 2D live imaging, and the tissues were finally imaged by PET/CT using the same protocol for live mice. To quantify total accumulated dose from fluorescence images, regions of interest were created by segmentation around each tissue to calculate total radiant efficiency in photons per second per square centimeter per steradian (p s⁻¹ cm⁻² sr⁻¹).

GWC Measurements.

Dissected tissues were collected, weighed, and measured using a gamma well counter (PerkinElmer). Each tissue was counted for 1 min to calculate the absolute iodine radioactivity/dose as counts per minute (CPM). After correcting for background activity, the CPM data were further corrected via decay coefficients based on the half-life of radioactive iodine-124 (4.2 days) or iodine-125 (59.4 days). Standard calibration curves of dose and CPM were used to calculate the dose of the radioisotope in each tissue (Supporting Figure S4). The injected dose percentage (% ID) of tissue was calculated as the dose in each tissue divided by the total dose injected. The injected dose percentage per gram of tissue (% ID g⁻¹) was determined by dividing the injected dose percentage (% ID) by the weight of the tissue.

Imaging Protocol for Solutions and Tissue Homogenates.

QD800 or ZW800 was diluted to different concentrations in PBS or in freshly prepared homogenates of VAT or liver in wells of a black well plate. The plate was imaged via a sequence of 30 excitation and emission filter pairs using the IVIS-Spectrum-CT instrument. Images of each well were processed by spectral unmixing to distinguish the spectra of QD800 or ZW800.

In Situ Fixed Tissue Fluorescence Microscopy.

Fixed tissues were imaged using an inverted Olympus IX-71 microscope. A 1.5 W fiber-coupled 750 nm laser (LDX Optrontics) was reflected off a dichroic (Semrock FF408/504/581/667/762-Di01-25×36) and passed through a 100× 1.49 NA oil immersion Nikon lens. Fluorescence was collected through a 785/40 nm emission filter (Chroma ET810/90m) by an EMCCD (Andor iXon DU-897E).

In Situ Live Tissue Fluorescence Microscopy.

Thick adipose tissues were imaged using a spinning disk confocal microscope (CSU-22, Yokogawa) built on an inverted Olympus IX-71 microscope. For maximizing excitation efficiency, lasers were aligned through the side port instead of using an optical fiber. A 100 mW 640 nm laser and 50 mW 405 nm laser (Coherent, OBIS) were used for exciting QD800 and Hoechst, respectively. A 100× 1.40 NA oil immersion Olympus lens was used to excite the sample and to collect emitted photons. Fluorescence emissions from QD800 and Hoechst were filtered through a 785/40 nm emission filter (Chroma ET810/90m) and a 480/40 emission filter, respectively. Images were acquired by an EMCCD camera (Andor iXon DU-897E). A 3D z-stack image was generated by scanning a z-axis piezo stage across 20 μm with 100 nm steps.

Statistical Analysis.

Data were presented as mean ± standard error (SE) unless indicated otherwise.