The choice of a suitable dye for in vivo imaging depends on many factors. The most important consideration would be the molar extinction coefficient and quantum yield of the dye in the NIR region [16,17]. Of all the dyes, cyanine makes up the majority of commercial fluorescent probes for in vivo applications. Cyanine is a synthetic dye family of the polymethine group with a conjugated chain of odd number of carbon atoms linked between two nitrogen centers (Figure 4) [18]. Among this class of compounds, carbocyanine dyes with indolic groups are readily available commercially and have also been synthesized in a large variety of analogues. The structure of carbocyanine is formed by reaction of two indolic isomers (either identical or different) linked on each end of a C1, C3 or C5 methine. The absorption and fluorescence emission wavelength is determined both by the chain length of the methine and the side chains attached to the indolic groups. Table 2 shows a series of commercially available carbocyanine dyes emitting in the visible to NIR range. These dyes generally exhibit high molar extinction coefficients but only have moderate fluorescence quantum yields up to 30%. A prominent representative of the NIR cyanine dye is Indocyanine green (ICG). ICG was first synthesized in the fifties [19] and is the only clinically approved dye available commercially. It has been used in optical imaging of changes in blood-brain barrier permeability after thrombus formation in a mouse model of cerebral venous thrombosis [20] and liver perfusion in mouse [21]. Clinically, ICG has been used in clinical retinal angiography [22], hepatic function testing [23] and imaging of human brain [24,25]. Since ICG binds tightly to plasma proteins and becomes confined to the vascular system, they are widely used for intraoperative assessment of vascular flow in cardiovascular surgery [26,27,28]. Because of its amphophilicity and a shortage of functional groups available for conjugation, ICG could only function as a non-specific probe for in vivo imaging.

Covalent attachment of targeting moieties such as antibodies, antibody fragments, proteins and peptides to fluorescent dyes can significantly improve the target-to-background signal. The strategy of conjugating antibodies to cyanine dyes were first demonstrated by Folli et al. and Ballou et al. [29,30] and applied to fluorescence imaging of tumor in mice. Antibodies, however, have found limited utility due to their unfavorable pharmacokinetics. The typical circulating half-life of antibodies is much shorter than the time required to access a small tumor with reasonable accumulation, thus rendering effective targeting rare using antibodies [31].

To improve the pharmacokinetics of in vivo contrast agents, a possible approach is to reduce the conjugate molecular size while still preserving the targeting affinity of the labeled dyes. Neri et al. demonstrated this idea by conjugating antibody single chain fragments selected from phage display libraries against an angiogenesis-associated oncofetal fibronectin isoform to cyanine dyes and applied to imaging of angiogenesis in a variety of animal models [32,33]. Alternative strategy is also shown in the work by Achilefu et al. and Licha et al. highlighting the successful application of receptor-specific peptide-dye conjugates for fluorescent imaging of tumors [34,35,36]. As a whole, fluorescent dyes can be readily functionalized to achieve targeting.

The concept of activatable NIR probes was introduced by Weissleder et al. for in vivo imaging of enzyme activity [37]. Enzyme-activatable probes contain either two identical or different fluorophores linked in close proximity to each other by a specific peptide linker. The fluorescence of the probes is essentially undetectable in the quenched state. After enzymatic cleavage, the fluorophores are separated to restore the fluorescence emission. A large number of activatable probes have been documented, among them the enzyme activation of NIR targeting dyes is mediated mainly by tumor associated proteases, such as cathepsins, caspases and matrix metalloproteinases [38,39,40,41,42,43]. The ability of activatable fluorescent probes to detect gene expression is particularly useful as a means to diagnose malignant molecular process in the early disease state [44].

Fundamentally, quantum dots (QDs) are semiconductor nanocrystals that absorb photons of light and re-emit photons at a different wavelength (Figure 5a) [54]. Typically, QDs are of ~2–20 nm in diameter depending on the core composition and the surface coating or functionalization. Most of the QDs reported have a core/shell structure (Figure 5b) with the core composed of atoms from periodic groups II–VI (CdSe, CdTe, CdS, PbSe, ZnS and ZnSe), III–V (GaAs, GaN, InP, InAs), and IV–VI (PbS). QDs differ from traditional organic fluorescent dyes and naturally fluorescent proteins in several important aspects. Firstly, QDs are extremely efficient in generating fluorescence, often by an order of magnitude higher than traditional fluorophores. In addition, QDs fluoresce without involving conjugated double-bond systems, thus exhibit greater photostability to enable long-term imaging without the concern of photo-induced deterioration. Another practical advantage of QDs is that their emission spectra are narrow and symmetric, thus minimizing any overlapping of colors in multi-component imaging applications. With the same underlying material, varying the size of the QDs could give rise to different emission peaks, giving QDs unprecedented tunability. This enables the use of a single laser to excite and achieve multicolor emission for multiplexed assays [53,55,56,57]. These unique properties of QDs have been increasingly exploited as bioimaging agents [58,59,60,61], theranostic platforms to deliver and track siRNA-based therapy [62,63,64,65], nanosensors to study intracellular trafficking and unpacking of DNA nanocomplexes [66,67,68,69], and cell lineage-tracing agents to follow the development of stem and progenitor cells in vivo developmental [70,71].

For in vivo imaging, the fluorescent emission wavelengths of the QDs ideally should be around 700–1000 nm, in the NIR region to minimize the endogenous fluorescence and the interference from major absorbers in the body. As presented in Figure 5a, CdTe [72,73], CdTe/CdSe [74], CdHgTe/ZnS [75,76], InAs [77], InAs_xP_1−x/InP/ZnSe [78], and PbSe [79,80] are the possible QDs that emit in the NIR I region. All these QDs are usually passivated with a layer of ZnS or ZnSe to protect the core from oxidation and to prevent the leaching of toxic heavy metal ions such as Cd, Hg, As and Pb into the surrounding solution. However, with the several reports indicating Cd-containing QDs showing cytotoxicity under extreme conditions [81,82,83,84,85], there is a rising concern of using them for long term in vivo applications. Therefore, attention has been channeled to develop new class of quantum dots comprising of less-toxic elements. Some of these newly developed NIR QDs include Cu-doped InP/ZnSe [86], CuInSe [87,88], and CuInS₂/ZnS [12,89,90].

Among the synthetic routes of QDs, the most commonly used technique is to first generate the core by injecting liquid precursors into nonpolar coordinating organic solvent at a temperature as high as 300 °C and subsequently growing the shell layer of ZnS onto the core [54,56,91,92]. QDs generated by this method are by default hydrophobic and are unsuitable for use in biological systems. To render them dispersible in aqueous medium, the QDs such as CuInS₂/ZnS are transferred into aqueous phase either by ligand exchange with dihydrolipoic acid/polyethylene glycol 1000 (DHLAPEG100) or by encapsulation in phospholipid micelles [60,90,93,94]. The resulting water-soluble QDs emit at wavelength greater than 700 nm and could achieve good quantum yield of 30%. In vivo sentinel lymph node imaging with the CuInS₂/ZnS QDs were performed in the work of Pons et al. by subcutaneous injection into regional lymph nodes (LN) of healthy mice. It was observed that the QDs preferentially accumulated at the injection site, in the two regional LNs and to a lesser extent in other organs over a 10-day observation period [95]. The results suggest that the LNs would be more sensitive toward acute QD toxicity. They further compared the inflammatory response of the two regional LNs treated with CuInS₂/ZnS and CdTeSe/CdZnS QDs. Histological sections indicated inflammation only occurred at 10 times dose concentration for cadmium-free QDs than for the Cd-containing counterparts (Figure 5c,d). This study highlights the advantage of cadmium-free QDs for safer NIR in vivo imaging.

Silicon is one element that is most widely used in the microelectronics industry due to its outstanding performance as an electronic material [96,97]. Its inherit nontoxic and environmentally friendly characteristic has attracted much attention to develop silicon quantum dots (Si QDs) for biological applications [98,99,100,101]. An important step toward realizing these applications is the fabrication of colloidally and optically stable, water-dispersible Si QDs. Silicon QDs are typically prepared in a two-step process: first by CO₂ laser pyrolysis of silane to generate non-photoluminescent crystalline Si particles with an average diameter of 5 nm [102]. Etching with mixture of hydrofluoric acid and nitric acid is then followed to reduce the size and passivate the surface of the particles [103]. Using this method Si QDs with size ranging from 1 to 5 nm and emission from blue to NIR can be obtained.

In 2004, Li et al. first demonstrated the potential of Si QDs for biological applications. They grafted water-soluble polyacrylic acid Si QDs to label Chinese hamster ovary (CHO) cells for in vitro imaging [104]. In the recent work by Park et al. they delivered Si QDs loaded with anti-cancer drug doxorubicin D-LPSiNPs into BALB/c mice and monitored both accumulation and degradation in vivo (Figure 6) [99]. Si QDs are exhibiting the attractive attributes to be an effective and biocompatible nanomaterials system for theranostic purpose; they will undoubtedly see intense effort for further improvement and innovative applications.

Fluorescence of noble metal nanoclusters (NCs) of gold and silver comprising of several to tens of atoms have drawn great attention in the past decade because of their remarkable optical properties [105,106,107,108,109,110,111]. Because of their minuscule size, high chemical stability and biocompatibility, NCs are increasingly being exploited for sensing [112,113,114], biolabeling [115,116] and bioimaging [117]. Typically, the size of NCs is less than 2 nm and their properties are governed by their subnanometer dimensions. Their size regime is intermediate of atomic and nanoparticles in which they no longer have plasmon resonance and Mie’s theory is non-applicable [111]. In fact, the size of noble metal NCs is comparable to the Fermi wavelength of the conduction electrons, leading to molecule-like features such as discrete size-dependent electronic transitions, fluorescence and charging properties.

Multitudes of strategies have been developed for the synthesis of gold nanoclusters (AuNCs). One of which is chemical reduction of Au precursors in the presence of thiol stabilizers to afford AuNCs that fluoresce in the blue to near-IR regions [118,119,120,121,122]. However these NCs exhibit low quantum yields (QY) (0.001–0.1%). AuNCs can also be prepared via template-assisted synthesis within the cavities of dendrimers [111] with greater than 10% QY at the expense of fairly long reaction time of ~2 days along with the production of large nanoparticles (NPs) as byproduct. Another commonly used technique to prepare AuNCs is by ligand-induced core etching of metal NPs to yield blue-emission NCs [123]. Fine-tuning of the core etching method led Lin et al. to generate NIR AuNCs by first producing 6-nm diameter gold nanoparticles core stabilized with dihydrogenlipoic acid (DHLA) (AuNC@DHLA) [124,125]. Conjugation of AuNC@DHLA with biomolecules can easily be achieved with the carboxyl group on DHLA via carbodiimide chemistry. Polyethylene glycol (PEG, 5 kDa), PEG-biotin and avidin have been attached in this manner. The shortcoming of these NCs is the low quantum yield of 1–3%. Nevertheless, the authors demonstrated the specific labeling capability of the AuNCs toward fixed human hepatoma cells (HepG₂) and also the non-specific uptake of the AuNCs by human aortic endothelial cells. The results make obvious that these AuNCs are relatively nontoxic.

With increasing interest in applying NIR fluorescent probes for bioimaging, effort has been devoted to developing biocompatible AuNCs via aqueous synthetic routes with biological molecules [126]. Xie et al. [127] has developed an innovative method by exploiting the reducing capability of bovine serum albumin (BSA) to prepare AuNCs consisted of 25 gold atoms with red emission [127]. The synthesis is similar to the biomineralization activities of organisms. Once added to aqueous BSA solution the Au ions were sequestered and entrapped within the protein molecules. The reduction ability of BSA was activated when the pH of the reaction was adjusted to ~12 and the ions were progressive reduced to AuNCs in situ (Figure 7a). Additional reductant, such as NaBH₄ would be required to yield AuNCs that emit at 683 nm. The quantum yield of the NIR AuNCs produced by this method was also low (0.1%). Despite that, Wu et al. demonstrated the applicability of these AuNCs for in vivo tumor fluorescence imaging using MSD-MB-45 and HeLa tumor xenograft models. They attributed the high accumulation of the NIR AuNCs in the tumor areas to the enhanced permeability and retention (EPR) effects (Figure 7b) [128].

Carbon nanotubes (CNTs) represent another class of nanomaterials attract intense interest since their discovery in 1991 [129]. CNTs are cylindrical tubes of sp² carbon with superior optical, electrical, and mechanical properties, thus have wide range of potential applications as nanoelectronics devices and engineered composite materials [48,49]. In addition, CNTs have been extensively explored as intracellular delivery vehicles for drugs, proteins and genes in cancer therapy [130,131].

CNTs are categorized as single-walled nanotubes (SWNTs) and multi-walled nanotubes (MWNTs). So far, only SWNTs have been reported to possess high emission in the NIR II region, making them suitable for deep-tissue in vivo imaging [132]. In general, SWNTs are most commonly synthesized by chemical vapor deposition that involves the heating of an alumina substrate with a layer of transition-metal catalytic nanoparticles, which serve as seeds to nucleate the growth of the nanotubes in a furnace with a stream of hydrocarbon gas flowing through the tube reactor [133]. SWNTs synthesized by this method have vast possibilities in the type of carbon tube “molecules”, giving rise to variation in band gaps ranging from ~10 meV to ~0.5 eV [133].

Welsher et al. suggested that the biggest obstacle in realizing the potential of SWNTs as NIR II probes is producing SWNTs with high quantum efficiency and good biocompatibility [132]. To overcome this issue, the authors came up with a strategy of first debundling and solubilizing the hydrophobic, pristine SWNTs in sodium cholate by sonification, followed by surfactant exchange to displace the sodium cholate with pegylated phospholipid (DSPE-mPEG). The resulting water-soluble SWNTs exhibit several emission peaks across the NIR II region when excited at 808 nm (Figure 8). When delivered into athymic nude mice via tail-vein injection, the systemic circulation and biodistribution of the SWNTs can be followed for over two minutes (Figure 8) [134]. This work shows the advantages of fluorescence imaging in the NIR II region and their potential impact in advancing nanomedicine.