Biosensing with Fluorescent Carbon Nanotubes

1Future challenges in medicine such as early disease detection, point‐of‐care diagnostics, and tailored therapies require novel methods of biosensing. Additionally, biosensors can provide insights into the complex dynamics of biological and chemical systems. Consequently, they are essential tools for both fundamental research and biomedicine. In particular, optical sensing approaches possess a great potential for contactless real‐time readouts that are required in biomedical research, as well as industrial healthcare and agriculture applications. During the last decade, the field of biosensors based on nanomaterials has seen vast improvements. These materials include carbon‐based nanomaterials such as graphene, graphene quantum dots, and carbon nanotubes (CNTs). Here, single‐walled carbon nanotubes (SWCNTs) are of particular interest. Their optoelectronic properties are sensitive to the surrounding environment, which makes them suitable for highly selective biosensing. When dispersed in aqueous solutions, SWCNTs fluoresce without bleaching in the near‐infrared region (NIR, around λ=870–2400 nm). This region of the electromagnetic spectrum is beneficial for detection and imaging as it offers an ultralow background and high penetration depths in biological tissues (tissue transparency window). Fluorescence methods using common visible fluorophores often suffer from high scattering, absorption, and autofluorescence, which limits the penetration depth and signal to noise ratios. Additionally, phototoxicity is increased by excitation of common fluorophores with visible (Vis) or ultraviolet (UV) light. Consequently, SWCNTs offer an advantage as they combine the biocompatibility and photostability required for optical sensing and imaging with emission in the NIR region. Furthermore, the structural diversity of SWCNTs promises tunable emission wavelengths. SWCNTs are highly sensitive to environmental changes, which is the basis for molecular recognition and was pioneered by optical sensors for glucose detection and DNA polymorphism. Both covalent or noncovalent functionalization approaches play an essential role in tailoring molecular interactions close to the SWCNT surface. By using such concepts, SWCNT‐based biosensors for many highly important biomolecules have been developed.

2More recently, this allowed chemical signaling to be mapped in a completely new manner, for example, release patterns of neurotransmitters from cells with high spatial and temporal resolution, which provides unique insights into fundamental biological questions. Moreover, recent advances have been made in remote in vivo biosensing applications by the multimodal optical detection of several analytes. By combining multiple nanosensor elements and integrating them into functional arrays, analytes can be identified and distinguished on the basis of their characteristic image signatures. Such a combination of optical nanosensors could pave the way for the next generation of fast and reliable in situ diagnostics. In addition, these approaches provide completely new opportunities for standoff process controlling, for example, fabrication of antibodies or monitoring in food and agriculture industries (smart plant sensors).

3In this Review we focus on optical biosensing with SWCNTs to give an update on this fast‐evolving field. We evaluate in detail the specificity, sensitivity, spatial resolution, and biocompatibility of different SWCNT‐based biosensors. This Review follows on from previous reviews, and discusses new chemical strategies developed in the last few years. SWCNTs can also serve as NIR labels. However, this is not discussed here and we refer to other excellent reviews.

4In Section 2, the basic structural properties and photophysics of SWCNTs as well as functionalization strategies are described. To conclude this section, we touch on the most important aspects of SWCNT biocompatibility. Section 3 contains an overview of general chemical recognition strategies. We provide a detailed and up‐to‐date summary of all currently accessible biomolecular target groups, including reactive oxygen species (ROS), neurotransmitters, proteins, antibodies, lipids, and sugars. This overview is complemented with mechanistic insights into how these sensors work. Finally, we provide a perspective on the field (Sections 3 and 4) and discuss possible future directions. This includes novel biological topics such as plants, advanced chemical tools (defects), methods for improved (hyperspectral) imaging, novel screening approaches, and multiplexing.

5Since the report of their structure, CNTs have attracted wide interest within the scientific community and beyond. Their remarkable mechanical, electrical, and photophysical properties have paved the way for applications in the fields of advanced materials, microelectronics, biosensing, imaging, drug delivery, and many more. Here, we will briefly describe the structure and photophysics of SWCNTs, followed by approaches to tailor their surface chemistry and biocompatibility.

6CNTs can be conceptualized as rolled‐up cylinders of graphene. Their properties are determined by the exact sp2‐hybridized carbon lattice as well as by the number of cylinders that are stacked into each other. CNTs derived from a single graphene cylinder are called single‐walled carbon nanotubes (SWCNTs), whereas tubes consisting of multiple layers are called multiwalled carbon nanotubes (MWCNTs). SWCNTs are commonly labeled using the chiral index (n,m), where n and m are integers that describe the carbon lattice structure (Figure 1 a). In this notation, the SWCNT is conceptually rolled up along the vector c=na1+ma2 (a1 and a2 are the graphene lattice vectors). Consequently, the roll‐up vector also determines the diameter. For SWCNTs, the reported diameters range from 0.4 nm to 10 nm.

7The roll‐up vector affects the density and energy of the electronic states of SWCNTs and consequently the optoelectronic properties are directly related to chirality. As a result, for n−m=0 (armchair configuration), SWCNTs are metallic, for n−m=3j (j∈ℕ∖0 ), semimetallic, and semiconducting for all other (n,m) chiralities. When SWCNTs are excited with light, an electron–hole pair (exciton) can be created and diffuses along the SWCNT axis. For semiconducting SWCNTs, the absorption of photons with energies corresponding to the visible spectrum of light typically leads to excitation to the second conducting band (Figure 1 b).

8Fast decay (femtosecond time scale) to the first conduction band followed by radiative recombination, then causes fluorescent emission in the NIR region (>870 nm), a region that is particularly interesting for biological imaging (Figure 1 c). Quantum chemical considerations predict 4 singlet and 12 triplet excitonic states. However, only the transition from the singlet state is optically allowed. As the energy of this state is higher than the majority of other singlet and triplet states, a variety of dark exciton decay pathways exist. For (6,5)‐SWCNTs, the size of an exciton is approximately 2 nm. During their lifetimes they diffuse in the range of 100 nm along the SWCNT axis. As all carbon atoms are located on the surface of the SWCNT, excitons are affected by the nanotube corona (i.e. the organic phase around the SWCNTs). Consequently, the photophysics of SWCNTs are highly influenced by chemical processes around their surfaces. This renders them ideal building blocks and transducers for chemical and biological sensing.

9The extended π‐system makes SWCNTs hydrophobic and consequently they easily aggregate in solvents like water. Therefore, an important step in the preparation of SWCNT‐based sensors is their functionalization to isolate, solubilize and colloidally stabilize single SWCNTs. The functionalization also serves the purpose to a) interact (specifically) with other molecules and b) translate this interaction into a fluorescence change.

10In the past years, different covalent and noncovalent modification strategies (Figure 2) have been developed. For a complete overview, we refer the reader to several excellent reviews and discuss only concepts relevant for sensing here.

11On a more abstract level, two strategies to assemble selective SWCNT‐based sensors have been used, namely screening and rational design (Figures 3 and 5). The first one relies on permutations of the organic corona around the SWCNT (e.g. deoxyribonucleic acid (DNA) sequence) whereas the second one uses known recognition motifs (e.g. antibodies).

12Noncovalent functionalization in aqueous solution is achieved by sonication with surfactants that form micellar structures around the SWCNT or through strong π‐π interactions with the SWCNT surface (Figure 2). Prominent examples of surfactants are sodium dodecylsulfonate (SDS), sodium dodecylbenzenesulfonate (SDBS), sodium cholate (SC), sodium deoxycholate (DOC), lithium dodecyl sulfate, Triton X‐100, and pluronic F127. Additionally, functional surfactants—for example, with a perylene core together with a hydrophilic dendron—adsorb through π‐π stacking and enable energy transfer. In general, a surfactant concentration above the critical micelle concentration is required to stabilize dispersed SWCNTs in solution. Thus, these approaches are limited with regards to experiments in complex (biological) systems.

13In contrast, functionalization with larger biopolymers enables the formation of stable conjugates. Here, DNA and ribonucleic acid (RNA) form strong π‐stacking interactions between the nucleobases and the SWCNT surface, thereby exposing their negatively charged phosphate backbones and solvating the SWCNT–nucleic acid complex (Figure 2). As the conformation of the SWCNT–nucleic acid complex is affected by changes in the local ion concentration, locked nucleic acids have been used as more rigid synthetic derivatives at higher salt concentrations.

14As alternative to nucleic acids, certain polycyclic aromatic compounds carrying hydrophilic moieties have effectively solubilized SWCNTs. Similar to the π‐stacking of those compounds, the functionalization of SWCNTs with peptides, proteins, and other polymers has been widely demonstrated (Figure 2). SWCNT‐based biosensors have been rationally designed by the attachment of antibodies (or analogues; Figure 4 a,b,d) and peptides (Figure 4 e) to polymers or by the adsorption of boronic acids (Figure 4 c) and aptamers (Figure 4 f) on SWCNTs. In cases when sonication would destroy the structural integrity of the (bio‐)polymers, primary suspension of the SWCNTs in a surfactant, followed by subsequent exchange to the polymer by dialysis has been employed. An alternative to this rational design is the screening/search for novel organic phases. This concept was named corona‐phase molecular recognition (CoPhMoRe). Here, a heteropolymer adsorbs onto the carbon nanotube surface and forms a new structure (corona) that serves as a molecular recognition site for interaction with an analyte. The biomolecules used are typically amphiphilic with hydrophobic domains that enable SWCNT adsorption and hydrophilic domains to be responsible for the entropic stabilization of the SWCNT in suspension and formation of a binding site for the analyte. It is important to note that the biomolecules/polymers alone do not necessarily need to interact selectively with the analyte of interest. As such, the formation of these recognition sites cannot be predicted and are typically found by screening or high‐throughput approaches. Prominent examples of CoPhMoRe screenings are the identification of SWCNT‐based neurotransmitter sensors as well as the adaptation of the CoPhMoRe concept to proteins (Figure 5).

15The covalent functionalization of SWCNTs introduces new σ‐bonds into the sp2‐hybridized SWCNT structure. In contrast to noncovalent functionalization methods, the conjugates promise higher stability. However, the uncontrolled introduction of covalent sp3 bonds (defects) destroys the electronic and optical properties and diminishes the intrinsic NIR fluorescence. One strategy to overcome this problem preserves the sp2‐hybridized structure of SWCNTs during their covalent functionalization. In contrast, a certain number of sp3 defects give rise to novel properties, such as red‐shifted emission features that are capable of single photon emission. Therefore, these defects are also called quantum defects or quantum color centers.

16These sp3 defects have, at low densities, been shown to increase the fluorescence of SWCNTs. Incorporation of these defects at low concentrations leads to the trapping of excitons and an alternative decay pathway (E11 *) that results in a new red‐shifted fluorescence feature (Figure 6 a). A wide range of sp3 defects has been incorporated into SWCNTs to increase the fluorescent properties by using diazo ether, aryl halide, (bis‐)diazonium, as well as Billup‐Birch and alkyl halide reductions. Additionally, O‐doping approaches using ozone and light, sodium hypochlorite, as well as hydroperoxides of polyunsaturated fatty acids have been reported to increase the red‐shifted emission of SWCNTs at low defect concentrations (Figure 2).

17Covalent functionalization approaches also offer opportunities beyond changes in the photophysics. Defects that can be further functionalized enable the conjugation of important biomolecules. Recently, maleimide defects were used to link proteins such as nanobodies and phenylalanine defects to grow peptides directly on the SWCNT surface, similar to a solid‐phase peptide synthesis (Figure 6 b). The covalent conjugation approach with dichlorotriazine allows subsequent nucleophilic aromatic substitution of the chlorides using amine‐containing linkers (Figure 2). Another example are defects that are already able to interact with other biomolecules, such as phenyl boronic acids that interact with saccharides, and change the E11 * (S11 *) and E11 (S11) emissions.

18It is interesting to note that the resulting bathochromic shifts (E11 *) caused by sp3 defects can be tuned using the electronic properties of the incorporated moieties. That said, defects provide a rich chemical playground and interested readers are referred to several excellent reviews. With regards to SWCNTs with aryl defects, electron‐withdrawing substituents generally introduce red‐shifts to the E11 * emission that can be correlated to the Hammet constants (σ) of the substituents. Furthermore, the E11 * red‐shift shows a 1/d 2 dependence on the diameter (d) of the SWCNT. The protonation of diethylamino‐substituted aryl defects (σHR2N+=+0.82 vs. σR2N=−0.66) is a good example of the effect caused by the inductive effects of substituents. Furthermore, this type of defects allows a precise sensing of the pH value down to 0.2 units through the changes in the E11 * emission (Figure 6 c). Apart from the introduction of defects for sensing, the covalent functionalization of SWCNTs can be used for site assembly using different (bio‐)polymers and linkers.

19Defects also change the exciton decay pathways and, therefore, affect the photophysics and sensing mechanism of SWCNT‐based sensors. This approach was recently used to perturb the sensing and elucidate the rate constants that are involved (Figure 6 d).

20In this study it was also found that a small number of defects can inverse the sensing response from a strong increase to a strong decrease in fluorescence.

21Biocompatibility is highly important for materials in direct contact to biological matter. Even though many biocompatibility studies exist, the conclusions are difficult to compare. The main reason is that different materials, surface reactions, and biological systems are compared, which leads to a noncoherent view. Moreover, the SWCNT field has evolved dramatically over the years and well‐defined chirality pure SWCNTs‐based sensors with ultrahigh purity are available today, whereas older studies used less well‐defined materials. In addition, the application itself determines the perspective. As a research tool, SWCNTs should not affect the biological system in such a way that the results are biased. For long‐term in vivo studies and applications in humans, the fate of SWCNTs in biological systems is highly relevant for their biocompatibility. SWCNTs have been shown to be susceptible to degradation by oxidative processes introduced by neutrophils and macrophages. Furthermore, the functionalization changes the surface properties of the nanotubes and, thus, ultimately the way SWCNTs interact with the molecules in a (biological) system.

22For example, endocytosis experiments have shown that the DNA sequence length plays an important role in endocytosis and retention time scales of DNA‐functionalized SWCNTs within mammalian cells. Additionally, experiments using a combination of NIR fluorescence spectroscopy and resonance Raman scattering have been used to analyze the fate of DNA‐functionalized SWCNTs through the endosomal process. Based on the experimental findings, the authors propose that DNA‐SWCNTs enter the cell, where they are transported into early endosomes. Maturation of the endosome begins with a decrease in the luminal pH value, which is followed by a series of physicochemical processes that transform the endosome into a lysosome, where the SWCNTs finally aggregate.

23As correct functionalization has been shown to alleviate the pathogenicity of SWCNTs, stable functionalization is one possible way to safeguard the future design of SWCNT‐based sensors in environments where long‐term stability is of the highest importance. Adequately functionalized SWCNTs have been shown to possess excellent biocompatible properties. A good example is the recently published long‐term biodistribution and compatibility assessment of DNA‐encapsulated SWCNTs after intravenous administration in mice. After an initial increase in the SWCNT fluorescence in the liver, the SWCNT fluorescence decreased rapidly over the course of 14 days. The same trend is also seen in the long‐term SWCNT biodistribution in different organs. By using hyperspectral microscopy, low levels of SWCNTs were detected in murine hearts, lungs, livers, kidney, and spleen tissues one month after injection. Assessment of these tissues after three and five months showed no SWCNT fluorescence in lung tissue, or in heart and lung tissues. Moreover, no abnormalities were found in chronically exposed tissues after hematoxylin and eosin (H&E) staining at all observed time points and the assessed biomarkers showed negligible changes up to four months, and minor changes after five months.

24The aforementioned studies suggest remarkable opportunities for SWCNTs in biomedical applications. As a consequence of the interplay between different materials, surface reactions, and biological systems it becomes evident that biofunctionalized SWCNTs represent a class of different materials. As for all new materials, the biocompatibility should be evaluated for every type of chirality, purity, functionalization, and route of administration. The scientific community is well‐aware of this problem and it has been pointed out that an assessment of these parameters in the context of biocompatibility depends on the context of the experiment, timescale, and the application of the nanomaterial. As a consequence, it is fundamentally important to place experimental data in the right context.

25An important requirement for a biocompatible design of SWCNT‐based sensors is an in‐depth understanding of the composition of the protein corona in biological media. In this regard, a recent study characterized the enrichment of certain proteins in the SWCNT corona.

26In the future, long‐term studies comparing the biocompatibility of different SWCNT subclasses (purity, chirality, surface chemistry) would be desirable to safeguard the development of biocompatible sensors. A foundation for the standardization of protocols could be the MIRIBEL (Minimum Information Reporting in Bio‐Nano Experimental Literature) reporting standard. As the functionalization of the SWCNT plays a critical role in the biocompatibility of SWCNT‐based sensors, the design of stable SWCNT functionalizations needs to be carefully ensured for long‐term applications. In particular, the recent advances in covalent functionalization strategies might, therefore, offer interesting starting points for future development.

27The discovery of band gap fluorescence from SWCNTs and their structure‐dependent NIR emission wavelength marks the starting point for SWCNT‐based sensors. Given the high surface to volume ratio of SWCNTs, it was quickly anticipated that SWCNT fluorescence would be sensitive to the chemical environment. The first generation of sensors targeted mainly smaller molecules including protons and reactive oxygen/nitrogen species (ROS/RNS). In these cases, the fluorescence changes were most likely caused by direct quenching. At the same time, it was discovered that different surface reactions with biopolymers lead to molecular interactions that are surprisingly specific even without using a standard approach with antibodies. This idea was conceptualized as corona‐phase molecular recognition (CoPhMoRe). During the last few years, great progress has been made in the chemical design of sensors for both biomedical and environmental applications. In the next sections we will give an overview on different sensing strategies, organized according to the molecular target. Here, sensing of the target molecule has partly environmental as well as biomedical applications. We focus on the advances in the last few years but also report previous studies (see Tables 1–, 9 in the Appendix).

28The ongoing advances in the development of recognition strategies have led to powerful biosensors based on SWCNTs. Various targets can be detected with high selectivity and sensitivity by combining a recognition unit with the SWCNTs. Recognition strategies (Figure 3) can mainly be categorized into a screening (Figure 5) and a rational approach (Figure 4). The first approach is in principle achieved with a library of synthetic organic phases (coronas) consisting of different amphiphilic polymers wrapped around SWCNTs and screened against a panel of various analytes to find a selective interaction. The latter approach is mostly applied for the detection of larger molecules such as proteins or sugars by conjugating a known binding partner of the target analyte to the SWCNT surface. Several approaches are based on the use of SWCNTs wrapped by single‐stranded DNA (ssDNA), whereby different lengths of the (GT) x sequence is probably the most used sequence up to now.

29It has shown its versatility in the detection of divalent ions, genotoxins, NO, H2O2, riboflavin, doxorubicin, β‐carotene, endolysosomal lipids, arsenite, and neurotransmitters, especially dopamine. This section highlights the major advances from the last few years for different categories of biomolecules. A detailed overview of most of the known fluorescent SWCNT‐based sensors, subdivided into the target categories, can be found in Tables 1–9 in the Appendix.

30ROS/RNS are important signaling molecules in many organisms, but their detection is challenging because they diffuse fast and have short lifetimes due to their high reactivity with O2 and other molecules. Since the finding of the first NO sensor based on SWCNTs the performance of SWCNT‐based ROS sensors has grown from the first selective detection of NO and H2O2 at the single molecule level and first in vivo applications to a new approach to study NO generation and spatiotemporal imaging of intracellular NO signaling. Recently, a mathematical model that calculated the NO concentration based on the change in the SWCNT fluorescence was derived. This was previously not possible due to a nonlinear fluorescence quenching rate in response to NO.

31ROS play a mediating role in the cell‐to‐cell communication of plants to activate defence mechanisms, whereby it has become clear that H2O2 is the primary mediator that responds to different stresses in plants. This has led to novel SWCNT sensor approaches to study ROS within plants. Wu et al. demonstrated remote H2O2 monitoring of plant health with sensitivity in the plant physiological range by using fluorescent SWCNT‐based sensors. Their rational approach was based on a DNA aptamer that specifically binds to the porphyrin hemin (HeAptDNA‐SWCNT). Hemin binds ferric ions, which undergo a Fenton‐like reaction with H2O2 to produce hydroxyl radicals (Figure 7 a) that directly quench the SWCNT fluorescence. For spatiotemporal in vivo monitoring, SWCNTs were embedded in leaves of plants and the plants exposed to different stresses such as UV‐B, high light intensities, and a pathogen‐associated peptide (flg22; Figure 7 b). The decrease in fluorescence reported remotely different aspects of the stress. These differences in fluorescence intensity quenching offer the possibility to interpret stress patterns in plants.

32Similar to this approach, Lew et al. developed a platform for H2O2 detection in leaves of different plant species. This sensor platform used a ratiometric approach, with (GT)15‐SWCNTs (G‐SWNT) that respond to H2O2 by quenching, possibly as a result of a charge‐transfer phenomenon, and (AT)15‐(6,5)‐SWCNTs (A‐SWNT) as an invariant reference (Figure 7 c). They infiltrated both G‐SWNT and A‐SWNT into spinach leaves and monitored the H2O2 signal with a standoff detection platform in real time (Figure 7 d). In the presence of different stresses, for example, tissue wounding, distinct waveform characteristics were observed (Figure 7 e), whereby the wave speeds in different plant species post‐wounding differed in the range of 0.44 to 3.10 cm min−1.

33In the same manner, Lew et al. developed a SWCNT‐based sensor system for the specific detection of arsenite in plants to monitor the uptake of the toxic heavy‐metal pollutant arsenic by using a self‐powered microfluidic system in real time. For this purpose, they infiltrated the sensors and the invariant reference into leaves of spinach, rice plants, and hyperaccumulating fern, which is able to preconcentrate and extract arsenic from soil (Figure 8 a). The intensity of the sensors increased steadily over several days, with the sensor response of the hyperaccumulating plant being significantly higher than those of the rice and spinach plants (Figure 8 b). Based on a kinetic model, the arsenite concentration in the leaf and the limit of detection (LOD) were calculated to be 4.7 nm and 1.6 nm as a function of the root fresh weight and uptake solution volume after 7 and 14 days (Figure 8 c). These examples show that SWCNT‐based H2O2 sensors are able to report plant stress on a microscopic and macroscopic level with potential applications in smart agriculture.

34Another macroscopic situation in which H2O2 plays an important role is wound healing. Safaee et al. developed a wearable optical microfibrous material with encapsulated SWCNT‐based sensors (Figure 8 d) to monitor the H2O2 concentration in wounds. Their approach was based on the ratiometric signal of (8,7)/(9,4)‐SWCNT chiralities, which differed in their response to H2O2 (Figure 8 e). The fluorescence signal was invariant to the excitation source distance, and exposure time, which enabled detection within commercial wound bandages with a wireless readout (Figure 8 f). These microfibers encapsulated the SWCNTs for at least 21 days without structural changes.

35Furthermore, Zheng et al. reported the selective interactions of SWCNTs coated with ten different ssDNA sequences in response to dissolved oxygen. The SWCNT emission intensity was quenched between 9 to 40 % depending on both the ssDNA sequence and SWCNT chirality in response to 1 atm O2 compared to samples purged with 1 atm argon, thus indicating that stronger coating interactions lead to reduced O2 access to the SWCNT surface. Since the quenching reversed completely after the removal of dissolved oxygen, it is probably based on physisorption on the SWCNT. Thus, the screening for fluorescence quenching by dissolved oxygen provides a simple approach to explore the structure‐selective interactions of ssDNA with SWCNTs.

36ROS can also be generated by enzymes and SWCNTs. Yaari et al. demonstrated the first SWCNT‐based sensor that reports the degree of enzymatic suicide inactivation. The approach was based on enzyme‐bound SWCNTs, which report fluorescence modulations by quenching and red‐shifting selectively in response to substrate‐mediated suicide inactivation of tyrosinase. Mechanistic insights revealed that the red‐shifted response is most likely a result of the generation of singlet oxygen during the enzymatic reaction, which leads to the binding of ssDNA on the SWCNT surface.

37Neurotransmitters are an important class of signaling molecules. To understand neuronal networks and linked neurological diseases, imaging with high selectivity and spatiotemporal resolution is necessary, which existing methods are currently not able to provide. In the last few years several SWCNT‐based sensors based on functionalization with DNA have been explored and it has been shown that sensitivity and selectivity depend on the exact DNA sequence.

38The first SWCNT‐based sensors for the detection of neurotransmitters were reported by Kruss et al. By using a screening approach, it was found that certain ssDNA‐SWCNTs change their fluorescence in the presence of catecholamine neurotransmitters such as dopamine. These sensors were reversible and showed sensitivities in the nanomolar range. Similar sensors were used for the high‐resolution imaging of cellular dopamine efflux from stimulated neuroprogenitor cells. Here, the sensors were immobilized on a collagen‐coated glass substrate to increase cell adhesion, and dopamine‐releasing neuroprogenitor (PC12) cells were cultured on top. In response to a stimulation event, the fluorescence intensity of the sensor layer image consisting of up to 20 000 pixels increased (Figure 9 a,b). This allowed both the spatial and temporal dynamics of dopamine release events to be studied with extraordinary high resolution and to identify hotspots (Figure 9 c,d).

39This approach of imaging many nanosensors under cells is also applicable to other neurotransmitters. Dinarvand et al. imaged the release of serotonin from human blood platelets in real time, as most of the serotonin is stored in blood platelets and not in the brain of humans. This serotonin sensor (NIRSer) consisted of a serotonin‐binding aptamer on a SWCNT, which exhibited an increased fluorescence emission of up to 80 % in response to serotonin (Figure 10 a). High‐resolution images of serotonin release patterns from single cells were obtained by placing the sensors below and around serotonin‐releasing cells (Figure 10 b,c). This approach allows serotonin release to be studied with unprecedented resolution and the time delay between stimulation and release to be resolved.

40In a similar fashion, artificially added serotonin was detected in acute brain slices by Jeong et al. In this case, a DNA sequence was found by an expanded screening approach from a library of around 6.9×1010 different ssDNA‐SWCNTs. These ssDNA‐SWCNTs exhibited a selective response to serotonin over serotonin analogues, metabolites, and receptor‐targeting drugs.

41The high spatiotemporal resolution of SWCNT‐based neurotransmitter sensors is especially useful when it comes to resolving parallel processes on the subcellular to cell‐network length scale. However, placing the sensors below cells can be a drawback, especially with cells that need to differentiate on this layer for several weeks. Elizarova et al. developed a new sensor paint approach (AndromeDA) to use sensors and study the dopaminergic signaling in primary neurons. In this case, the sensors were adsorbed (′painted′) onto the complex cell networks, which included different cell types. This approach allowed the heterogeneity of dopamine release events to be quantified from up to 100 release sites (varicosities), which is highly important to understand the information processing and plasticity of neurons.

42An effect that has to be accounted for in these studies is that SWCNT fluorescence is affected by changes in the local cation concentration, which is also a hallmark of neuronal activity. To circumvent this problem, Gillen et al. used locked nucleic acids to develop sensors with improved stability to cation‐induced fluctuations of the fluorescence intensity. By systematically introducing locked bases along the (GT)15‐DNA sequence they found that the fluorescence stability in the presence of Ca2+ ions depends on the type of the locked bases. Certain SWCNT chiralities exhibited improved stability against Ca2+ ions and retained their ability to detect dopamine in the presence of Ca2+ ions, thus highlighting the importance of the exact conformation of the nucleic acid sequence. Moreover, the detection of both Ca2+ and dopamine was possible by monitoring multiple chiralities simultaneous.

43Interestingly, the fluorescence responses of SWCNTs suspended in sodium cholate to dopamine and serotonin can be altered by modulating the exposed area by the surfactant concentration. However, such surfactants would not be compatible with cellular systems.

44A central challenge in biomedicine is the controlled delivery (uptake, transport, and release) of (nano)materials such as sensors and pharmaceuticals. Even whole cells can serve as vehicles to take up such materials. Certain immune cells (neutrophils) have been shown to be suitable for cargo delivery by hijacking a process known as neutrophil extracellular trap formation (NETosis). During NETosis, cells lyse after rupture of the cellular membrane by chromatin swelling. Meyer et al. showed that human immune cells take up ssDNA‐SWCNT‐based sensors and can be triggered to release the cargo after a certain time by using NETosis. Moreover, the sensors maintained their functionality to detect dopamine and H2O2, which offers opportunities for in vivo delivery.

45All these discussed neurotransmitter sensors responded by a fluorescence increase. Interestingly, the introduction of a small number of aryl defects into ssDNA‐functionalized SWCNTs completely reversed the sensing response (Figure 6 d). The E11 emission slightly decreased and strongly decreased the red‐shifted E11 * emission. Apart from new insights into the sensing, this approach enables ratiometric detection schemes. For a more detailed overview on the biological relevance of catecholamine neurotransmitters and alternative detection methods (e.g. electrochemical) readers are referred to the literature.

46Beyond neurotransmitters, recognition strategies for other small molecules have been developed, for example, adenosine 5′‐triphosphate, nitroaromatics, riboflavin, l‐thyroxine, oestradiol, doxorubicin, and steroids.

47Finding a specific molecular recognition element is difficult for many of these biomolecules, such as hormones, due to their chemical similarity. Therefore, Lee et al. used a polymer self‐templating synthetic approach, which is based on the attachment of a chemical appendage similar in molecular weight and structure to the target analyte to create a binding pocket within the corona. This approach reduced the library size for screening and led to implantable SWCNT‐based sensors for the selective detection of the human steroid hormones cortisol and progesterone.

48Recently, the first reversible fluorescent SWCNT‐based sensor for volatile organic compounds was also reported, which has potential for the detection of wine spoilage. For this, Shumeiko et al. used peptide‐encapsulated SWCNTs, which were adsorbed onto a polystyrene cuvette to detect low concentrations of acetic acid (down to 0.05 % (v/v)) in air. Using (6,5)‐SWCNTs, which fluoresce below 1000 nm they demonstrated the detection with a low‐cost Si‐based camera (Figure 11 a). The sensor was exposed to different concentrations of acetic acid, which quenched the fluorescence but was reversible when switching to clean air (Figure 11 b). The ability to identify wine spoilage was investigated by using two wine types with and without the addition of acetic acid to simulate an undesirably high acetic acid concentration (Figure 11 c).

49In an extension of this study, this system was expanded to an array of five different peptide‐encapsulated SWCNTs on a nitrocellulose paper. The optical patterns enabled the distinction of volatile molecules such as ethanol, methanol, and 2‐propanol as well as the aromas of red wine, beer, and vodka by linear discriminant analysis and machine learning.

50One way to detect or study small molecules is to mimic parts of larger biomolecules, such as enzymes. Dong et al. screened a library of 24 amphiphilic polymers to find a corona phase which demonstrates a binding specificity very similar to the enzyme phosphodiesterase type 5 (PDE5), which catalyzes the hydrolysis of secondary messengers. The SWCNT‐based sensor consisted of a poly(methacrylic acid‐co‐styrene) motif. This synthetic corona mimics the H loop of the native enzyme and is, thus, able to bind to Vardenafil, a PDE5 inhibitor, and its molecular variant as a result of the unique corona phase configuration. It is selective over other off‐target inhibitors, but not completely over the chemically similar inhibitor Sildenafil.

51One of the challenges in SWCNT‐based sensing is the heterogeneous material that is used for most sensors. Even though purification has made tremendous progress, getting access to chirality‐pure SWCNTs with a tailored surface chemistry has been a challenge. Nißler et al. showed sensing of small molecules such as riboflavin and ascorbic acid as well as pH value with chirality‐pure SWCNTs by using aqueous two‐phase extraction and a subsequent surface functionalization exchange process. The chirality‐pure sensors were up to ten‐times brighter than mixtures of SWCNT chiralities, and enabled insights into the impact of chirality and handedness of SWNCTs and the sensing mechanism. Additionally, long‐time stability over 14 days was demonstrated as well as ratiometric and multiplexed sensing based on the non‐overlapping fluorescence spectra (Figure 22 b,c).

52On a macroscopic level, monochiral SWCNTs were used by Nißler et al. to detect polyphenols in and around plants. Polyphenols are secondary metabolites and messenger molecules that are released from leaves and roots as a chemical defence against pathogens and herbivores (Figure 12 a). Certain polyethylene glycol phospholipids (PEG‐PL) were identified for the selective detection of different polyphenols over interfering molecules such as sugars or H2O2. The SWCNT‐based sensors responded through quenching and red‐shifting of up to 20 nm, for example, to tannic acid (TaA, Figure 12 b). To image the plant polyphenol release over time, the sensors were embedded in agar, soybean seedlings were plated on top, and polyphenol secretion was triggered with a pathogen‐derived elicitor, which resulted in a decrease in the NIR fluorescence over time (Figure 12 c). These sensors help to understand this plant defence mechanism and could improve the breeding of stress‐resistant plants for precision agriculture.

53The investigation of lipid‐linked diseases is challenging because methods for accurate in vivo monitoring of lipid accumulation have been missing. The Heller group was the first to address this issue by developing a SWCNT‐based optical sensor, which non‐invasively detects the lipid flux within the lumen of endolysosomal vesicles in vitro and in vivo. In a first approach, they used (GT)6‐functionalized (8,6)‐SWCNTs, which fluoresce at 1200 nm and responded through a wavelength shift to biological lipids and water‐soluble lipid analogues.

54By incubating the sensors with fibroblasts from a lysosomal storage disorder Niemann‐Pick‐type C patient in vitro, the sensors localized in the lumen of endolysosomal organelles without affecting their properties and resolved the lipid accumulation down to the subcellular level in real time. The authors proved the reversibility of this sensor by administering a drug which reverses the disease phenotype.

55The second approach was based on the screening of several ssDNA‐SWCNT chirality combinations to identify CTTC3TTC‐(9,4)‐SWCNTs with the greatest wavelength shift of up to 8 nm in response to lipid accumulations. The emission wavelength at 1125 nm is spectrally separated from the lipid absorption band at 1210 nm, thus facilitating promising in vivo applications. Molecular dynamics calculations led to the assumption that the lipid molecules sphingomyelin and cholesterol bind to the SWCNT surface by hydrophobic interactions, thereby decreasing the water density in the SWCNT environment and leading to the observed blue shift (Figure 13 a). To validate the sensor functionality in live cells, endolysosomal lipid accumulation was induced in macrophage cells with chemical inhibitors to mimic different lipid phenotypes. After sensor incubation, a blue‐shift was observed for all sensors within the drug‐treated cells compared to the control (Figure 13 b). For in vivo applications, two mouse models were used with lipid accumulation within the organelles of many cell types, for example, in Kupffer cells. By intravenously injecting the SWCNT‐based sensors, a rapid decrease of the SWCNT fluorescence and consequently removal of SWCNTs from all parts of the mice except for the liver were observed. The sensors were also able to report uptake and endolysosomal lipid accumulation of oxidized low‐density lipoprotein (oxLDL; Figure 13 c). Therefore, these types of sensors provide novel insights into the complexity of lipid metabolism and related health states.

56Proteins are one of the major biomacromolecules. Consequently, the study of protein–protein interactions helps to understand the function of proteins or to find novel drugs. As discussed above, the detection of proteins with SWCNTs is either based on attaching a known natural recognition element or a synthetic heteropolymer to the SWCNT surface (Figure 3). Such studies go beyond detecting the presence of a whole protein.

57Williams et al. developed the first label‐free optical sensor for the detection of glutathione‐S‐transferase (GST) fusion proteins based on glutathione‐(TAT)6‐SWCNTs (GSH‐DNA‐SWCNTs; Figure 14 a). The system used distinct differences in the emission wavelength and intensity of SWCNT chiralities in response to GST and, consequently, the ratiometric intensity of two chiralities (8,6)/(9,1) (Figure 14 b). The functionality of this sensor in response to four different GST fusion proteins was tested, which all showed a similar ratiometric response and a LOD in the low nanomolar regime. This approach has potential for the tracking of protein expressions in real time.

58The detection of insulin is essential for diabetes as it controls the glucose levels in blood. The first optical SWCNT‐based sensor approach for detecting insulin was realized in 2010 by Cha et al. They used an insulin‐binding aptamer, which showed a highly specific and sensitive decrease in the fluorescence intensity on forming a guanine quadruplex. Furthermore, they incorporated these sensors in a collagen extracellular matrix and demonstrated sensor reversibility with enzymatic proteolysis and the detection of insulin secreted by pancreatic β‐cells.

59Recently, insulin detection was enabled by screening a library of PEG‐conjugated lipids. It was found that the C16‐PEG(2000 Da)‐Ceramide‐SWCNT complex causes a significant decrease in the fluorescence intensity in the presence of insulin compared to other relevant proteins present in human whole blood. Additionally, nonspecific recognition mechanisms such as hydrophobicity or molecular weight were ruled out. For example, the sensor response to shorter insulin fragments was measured, however no correlation to the molecular weight was observed (Figure 14 c). The analysis of insulin fragments was further supported by enthalpy measurements that showed no affinity of the PEG‐conjugated lipid itself, which highlights that the exact CoPhMoRe phase bound to the SWCNT is responsible for the insulin binding. In more complex environments, such as serum, the sensor affinity was lower but still sensitive enough (Figure 14 d).

60The application of sensors in complex biological samples is often challenging. The spontaneous adsorption of proteins onto all kind of materials changes the actual corona structure and might affect sensing. To improve the performance of sensors in those protein‐rich environments, a fundamental understanding of the interaction between sensors and their biological environment is necessary. Pinals et al. addressed this issue by studying the protein corona formation on (GT)15‐SWCNTs in cerebrospinal fluid and blood plasma by mass spectroscopy.

61Their results showed strong binding to fibrinogen and other proteins involved in blood clotting, lipid transport, and complement activation. The identification of interactions responsible for the formation of protein corona revealed that the outer corona formation can be reduced by optimizing the electrostatic interactions through the sensor design and dynamic flow conditions (e.g. with lateral flow assays or microfluidic systems), while entropic calculations must be considered for the inner corona. This study highlights the urgent need to investigate sensors not only in simple buffers but biologically complex environments. Most recently, Ehrlich et al. developed a complementary approach using an insulin aptamer (found within the natural insulin gene promoter) functionalized to the SWCNT surface through a (AT)15 ssDNA anchor sequence. In contrast to the synthetic PEGylated‐lipid, which has no prior affinity to insulin, this aptamer possesses a known affinity to insulin. However, the observed sensitivity was lower than the previous approach.

62Most of the SWCNT‐based sensors that have been discussed so far were measured in solution. The immobilization of SWCNTs on different porous paper matrices, for example, nitrocellulose, has several advantages for a robust assay. Paper‐based immobilization enabled analyte detection within non‐aqueous solvents such as edible oil, which was previously not possible. To further extend this system the authors used wax to pattern hydrophobic regions onto the paper to create a multiplexed one‐dimensional sensor barcode consisting of different ssDNA‐wrapped SWCNTs.

63Another important class in biomedical diagnostics are antibodies. The detection of immunoglobulin G (IgG) with SWCNTs was realized by using chitosan‐wrapped SWCNT noncovalently modified with immunoglobulin‐binding proteins. Recently, Kozawa et al. designed a flexible fiber optic interface coupled to nanosensors that was capable of detecting the aggregation status of human IgG by reporting the relative fraction of monomers and dimer aggregates with sizes of 5.6 and 9.6 nm. For this purpose, the SWCNT‐based sensors were incorporated into a hydrogel (HG) and attached to the end of a fiber waveguide. Proteins are also part of pathogens and consequently disease markers. Pinals et al. developed a SWCNT‐based sensor which is functionalized with the angiotensin‐converting enzyme 2 (ACE2), a host protein which shows a high binding affinity for the SARS‐CoV‐2 spike protein (Figure 14 e). A twofold NIR fluorescence increase was detected 90 min after the addition of the purified spike protein (Figure 14 f). Passivation with a hydrophilic polymer was used to enable detection of the spike protein in saliva and viral transport medium.

64However, antibodies are not always available for all targets. In addition, the development of new recognition units can be expensive and tedious, which is why new approaches are directed to the development of multiplexed sensor arrays to overcome the limited selectivity of existing single sensors. Recently, Yaari et al. developed a SWCNT solution‐based sensor platform to detect multiple gynecologic cancer biomarkers in uterine lavage samples. The array consisted of eleven different ssDNA‐SWCNT sensors, and the optical change in the intensity and wavelength was extracted for twelve chiralities present in the sample, which resulted in 132 individual ssDNA‐SWCNT complexes. With machine learning algorithms a classification accuracy (F1 score) of 0.95 was achieved. With retraining, this sensor platform may not be limited to the detection of cancer biomarkers. The large variety of possible SWCNT chiralities in combination with unlimited SWCNT wrappings opens possibilities to meet the rising demand of new recognition strategies.

65Sugars are important building blocks and metabolites. Glucose, in particular, is a major target, for which continuous monitoring of the glucose level in blood is desired. SWCNT‐based sensing ranges from the use of glucose‐specific enzymes or proteins to the first affinity sensor based on the competitive binding between glucose and its polymer dextran. Although improvements were made, the first approaches suffered from limited reversibility and/or physiological detection range. One sensor that meets the requirements is based on the functionalization of SWCNTs with glucose oxidase (GOX), a glucose‐specific enzyme. The addition of glucose causes an increase in fluorescence emission. The proposed mechanism is based on SWCNT fluorescence being quenched by defect sites on the SWCNT surface, which are hole‐doped through oxygen adsorption. The addition of glucose causes an oxidation of the GOX wrapping, which behaves as an electron donor and passivates the oxygenated sites of the SWCNT, thereby resulting in a fluorescence increase. This effect is reversible by removing the glucose. The sensor showed responses to five other tested saccharides, but with the highest response to glucose.

66Another recognition element for saccharides is phenylboronic acids, which have been used to functionalize SWCNTs noncovalently for the detection of sugars. Recently, covalent aryl‐boronic acid defects were also incorporated in SWCNTs. Upon interaction with fructose and glucose, these sensors decreased in fluorescence intensity and the E11 * signal shifted, which can be used for spectrally encoded sensing.

67One of the most abundant and important types of biomacromolecules are nucleic acids that store and process genetic information. Using a construct of a complementary capture sequence connected to a (GT)15‐sequence serving as an anchor, thus providing colloidal stability, SWCNTs were recently used to detect hybridization events of microRNA and other oligonucleotides directly in serum, urine, and in mice in vivo. Upon the addition of complementary nucleic acids, a specific blue‐shift for different chiralities upon hybridization was observed. Additionally, the sensor response was reversible through toehold‐mediated strand displacement and the sensors possessed a LOD in the picomolar range.

68Further development of this sensor led to the first SWCNT‐based sensor for the detection of HIV in serum. Harvey et al. discovered that SDS‐denaturized serum proteins lead to an enhanced optical response of the SWCNTs in response to DNA hybridizations. They hypothesized that the addition of SDS ensured both the liberation of the viral RNA genome and the denaturation of the proteins which competitively bind to the freed surface of the sensor. The interaction leads to a blue‐shift in the SWCNT emission (Figure 15 a). This was first shown for hybridization with complementary target miR‐19 DNA compared to control R23 DNA (Figure 15 b). A dose‐dependent enhancement of the blue‐shift occurred in the region of the critical micelle concentration of SDS; thus, the denaturation of proteins by SDS is considered to involve an unfolding process of the tertiary structure of the protein to complete denaturation. As the SDS concentration was increased, the amount of denatured protein absorbed onto the SWCNT surface after hybridization of the DNA increased and saturated at 2 % SDS. For HIV detection, the recognition strategy was based on a sensor consisting of (GT)15‐T15‐SWCNTs, which hybridize with the polyadenylation elements of HIV RNA (Figure 15 c). In a similar way, a control sensor with the noncomplementary capture sequence, namely (GT)15‐A15‐DNA, was constructed. The wavelength shifts of both the sensor and negative control in the presence of HIV particles treated with 1 % SDS were recorded over time and and the sensor displaying the complementary capture sequence showed a blue‐shift of around 3 nm after 180 min.

69A completely different application of DNA chemistry on SWCNTs was established by Cha et al. Based on the consumption of chemical energy delivered by the RNA molecules, they developed a synthetic motor that transports nanoparticles through the mechanical motion of DNA conformation changes along the SWCNTs. Movements of the motor of over 3 μm with a speed of 1 nm min−1 were observed.

70So far, the detection of enzymes with SWCNTs has been shown for the characterization of proteases and DNases as well as cellulases and pectinases. The hydrolytic enzyme activity was measured by Kallmyer et al., who used hydrolytic enzyme wrapped SWCNTs that respond to the target enzyme with a fluorescent intensity quenching because of the degradation of the enzyme‐wrapping. Different polymer wrappings consisting of polysaccharides and polypeptides were used to study cellulase, pectinase, and bacterial protease. Most recently, they used this approach to evaluate the enzyme activity in soil using a low‐cost multiplexed and portable fluorimeter able to perform the measurement outside the laboratory only minutes after extraction from the field. As a consequence of the fresh nature of the soil sample, field tests indicated activities an order of magnitude larger than those obtained in benchtop experiments.

71Enzymes are also released by microorganisms, which can be used to fingerprint them. Nißler et al. chemically tailored SWCNTs to detect enzymes such as DNases and proteases. For targeting extracellular proteases, SWCNTs were modified with bovine serum albumin (BSA), which serves as an enzymatic substrate, while SWCNTs were functionalized with calf thymus (CT) DNA for reporting DNase I and S. aureus nuclease activity. The sensors showed a fluorescence decrease, most likely as a result of decomposition of the BSA surface coating. These SWCNT‐based sensors were further used for the discrimination of bacteria, which are known to alter their chemical environment through the release of signaling molecules, enzymes, and metabolites (see section below).

72In contrast, Shumeiko et al. used peptide‐encapsulated SWCNTs, which also responded through a fluorescence decrease upon enzymatic digestion of the SWCNT wrapping. They utilized a low‐cost paper‐based dipstick system, with which they evaluated the trypsin activity in urine samples as a mimic for acute pancreatitis, where abnormal trypsin concentrations are common.

73To study the enzyme myeloperoxidase, which is involved in the regulation of inflammation processes, He et al. used a ratiometric system based on graphene oxide (GO) wrapped SC‐SWCNT sensors and GO‐wrapped carboxymethylcellulose (CMC)‐SWCNTs as a reference. GO and SC‐SWCNTs showed an opposed fluorescence signal in response to enzymatic degradation. Whereas the blue fluorescence intensity of GO was increased due to oxidation and degradation of GO leading to the formation of graphene quantum dots, the NIR emission of SWCNTs decreased due to the generation of defects on the SWCNT surface. In contrast, the CMC‐SWCNT reference was almost stable as a result of the better surface protection of the CMC‐wrapping.

74Microbial infections are one of the major causes of mortality worldwide. Currently, the limited number of diagnostic methods in combination with increasing antibiotic resistances demonstrate the rising need for the rapid, contactless, and specific detection of pathogens such as bacteria. The optical properties of SWCNTs promise advantages for pathogen detection. Bardhan et al. developed M13 bacteriophage functionalized SWCNTs (M13‐SWNTs, Figure 16 a ), which are able to distinguish F′‐positive and F′‐negative bacterial strains by modulation of the fluorescence intensity. The M13 bacteriophage has a known binding affinity to F′‐positive E. coli strains. Therefore, they intramuscularly infected the right flank of living mice with E. coli strains, either F′‐negative DH5‐α strains (Figure 16 b, left) or targeted F′‐positive JM109 strains (Figure 16 b, right) and observed a 1.6‐fold intensity increase over the nonspecific DH5‐α strains. Injection of PBS into the right flank served as a control. To extend this SWCNT‐based label to a wider range of bacterial strains lacking F′‐pili, they additionally attached an antibacterial antibody on those M13‐SWCNTs (anti‐S. aureus‐M13‐SWNT) through a streptavidin–biotin reaction. This complex specifically detected S. aureus intramuscular infections in a mouse model in vivo (3.1‐fold intensity increase over the nontargeted M13‐SWCNTs, Figure 16 c).

75This approach is a NIR labeling rather than sensing. In contrast, Nißler et al. created a concept to remotely distinguish six important pathogens using an array of SWCNT‐based sensors: four for specific bacterial target detection based on a rational approach and four generic lower sensitivity sensors together with an invariant reference consisting of NIR fluorescent Egyptian Blue nanosheets (Figure 17 a).

76To overcome unspecific effects arising from the complex composition of bacterial media, the sensors were incorporated into a hydrogel (HG) array, in which the pore size was varied in accordance to the size of the analyte. For the detection of small molecules (such as siderophores), HGs with a low porosity were used, whereas HGs with a high porosity allowed large enzymes to diffuse to the sensors. The sensors were exposed to clinical isolates of bacteria and the unique change in the fluorescence intensity for each sensor was monitored remotely. Within 24–72 h, a unique fingerprint in response to the tested pathogens was visible (Figure 17 b), which further allowed differentiation of the pathogens by principal component analysis (PCA; Figure 17 c). Besides this spatially encoding, spectral multiplexing was implemented to differentiate S. aureus and P. aeruginosa by using two monochiral SWCNT‐based sensors with different wavelengths compared to the reference. Moreover, the signal of the sensor could be detected through tissue to a depth of >7 mm, which highlights the potential for biomedical in vivo applications.

77In SWCNT‐based sensors or probes, the SWCNTs serve as transducer elements that translate chemical changes caused by an analyte in the vicinity of the SWCNT into a fluorescent signal. Thus, SWCNT‐based sensing involves molecular recognition and signal transduction. The precise mechanisms are most likely different for different analytes and different surfaces. However, from the literature one can distinguish several possible generic mechanisms (Figure 18).

78Direct quenching, that is, the decrease of fluorescence, is caused by adsorption of the analyte onto the SWCNT surface. Changes in the pH value caused by the addition of an acid can cause protonation of the SWCNT sidewall and result in direct, reversible quenching of the SWCNT fluorescence (Figure 19 a). Mechanistically, this can be explained by the injection of an electron hole into the π‐system near the protonation site. Excitons encountering such an electron hole will be quenched through a nonradiative Auger process. Furthermore, electron transfer between the valence band of 3,4‐diaminophenyldextran‐functionalized SWCNTs and the lowest unoccupied molecular orbital (LUMO) of nitrogen oxide results in rapid, reversible quenching of the SWCNT fluorescence. Both of these interactions take part in the vicinity of the SWCNT and, consequently, solvent effects should play a large role.

79Within the dimensions of the exciton, the fluorescence of SWCNTs is very sensitive to the surrounding environment. To study the contributions of the solvent to the fluorescence of SWCNTs, fluorescence spectra were analyzed in different dielectric environments. By using the solvatochromic shifts, a semiempirical scaling model was developed that linked optical with structural parameters and suggested an inverse dependence of exciton polarizability on the diameter and the square of the transition energy. In nonpolar solvents, the solvatochromic modulation of the fluorescence intensity becomes more pronounced for larger diameters. Changes in the solvatochromic shift, on the other hand, were more pronounced for SWCNTs with smaller diameters.

80In general, the displacement of H2O or DNA from the surface of SWCNTs by surfactants leads to a strong blue‐shift and increase in the fluorescence intensity. Interestingly, the change in the fluorescence characteristics of pristine SWCNTs and DNA‐coated SWCNTs immobilized in gel are highly similar. This suggests that considerable portions of the nanotube surface are exposed to H2O. However, to date, a local model of solvatochromism that accounts for the nonhomogeneous structure around SWCNTs is missing.

81One of the best understood systems is the recognition of small molecules by DNA‐functionalized SWCNTs. Here, the mechanism of sensors for catecholamine neurotransmitters such as dopamine have been studied in greater detail (Figure 19 d,e).

82As these molecules are redox‐active, they could reduce or oxidize either the SWCNT or the surrounding organic phase, thereby affecting the fluorescent properties. To study this potential mechanism, the redox potential and the fluorescent response of certain analytes were correlated. This study showed that molecules with a negative redox potential are more likely to increase the SWCNT fluorescence (Figure 19 b). However, as molecules of the same redox potential can induce drastically different fluorescence responses, the redox potential alone cannot account for the fluorescence changes observed. Likewise, fluorescent responses of ssDNA‐functionalized SWCNTs to dopamine and riboflavin cannot be correlated to the amount of adsorbed nucleotides/DNA molecules on the SWCNT surface (Figure 19 c), which suggests that more complex conformational changes are responsible for the change in the SWCNT fluorescence. Molecular dynamics simulations showed that a stacking of dopamine with DNA‐functionalized SWCNTs leads to interactions between the phosphate backbone of the DNA as well as the hydroxy and amine groups of dopamine (Figure 19 d,e). As a result of this interaction, the phosphate backbone moves toward the SWCNT surface and the electrostatic potential at the SWCNT surface changes (Figure 19 d,e). It is also known that the diffusion coefficient of the excitons in surfactant‐containing systems changes with the surfactant identity and, furthermore, correlates with the fluorescence intensity.

83As pristine SWCNTs do not show any fluorescence response to dopamine and differently functionalized SWCNTs display different affinities to neurotransmitters, it follows that the organic phase (DNA) governs both the sensitivity as well as selectivity for this neurotransmitter. The used biopolymers are typically charged and, consequently, electrostatic interactions play an integral part for biosensing. It has been shown that the presence of certain salts alters the conformation of the DNA wrapping on the SWCNT and decreases the electrostatic repulsions between equally charged molecules. To reduce ion‐induced fluorescence effects, the flexibility of the DNA can be altered by using xeno nucleic acids. A good example of the influence of electrostatic repulsion and screening effects is the increased SWCNT surface accessibility at higher salt concentrations (Figure 19 f). In a surfactant‐containing system, the exposed surface is covered by the surfactant, which causes a blue‐shift of the NIR fluorescence, again demonstrating that the dynamics of the organic phase around the SWCNT is crucial for the sensing mechanism.

84By modulating the exposed surface area, it is furthermore possible to tune surfactant‐suspended SWCNTs to respond to different bio‐analytes. Together, these findings indicate that fluorescence is modulated by the precise 3D arrangement of the molecules, ions, and water molecules in the vicinity of the nanotube.

85As described above, the introduction of certain sp3 defects into the carbon lattice of SWCNTs can increase the NIR fluorescence through the trapping of excitons. The defined introduction of quantum defects, thus, provides a way to perturb the exciton decay and elucidate the involved processes that ultimately affect the fluorescent response. As the NIR fluorescent response of SWCNT‐based sensors to dopamine in H2O and D2O did not show major differences, electronic‐to‐vibrational energy transfer (EVET) seems not to be a main factor in (dopamine) sensing. In contrast, the correlation between the length of the SWCNTs and their fluorescent response seems to indicate that quenching at the ends plays a role. However, the fluorescence response was independent of the variation of defect density. Together with the finding that a small number of quantum defects reverse the fluorescent response of DNA‐functionalized SWCNTs to dopamine (Figure 6 d), it follows that multiple rate constants are affected by the analyte. Computationally, the experiments were best explained by a three rate constant model (3RC) that includes a decrease in the nonradiative decay from the E11 state (k nr), an increase in the exciton diffusion constant (De), and an increase in the nonradiative decay constant from the E11 * (k nr *) caused by dopamine. Together, these insights highlight the complex interplay between photophysics and molecular recognition as well as new avenues to tailor sensing using defects.

86The processes related to sensing happen on a certain time scale. How fast an analyte binds or dissociates from a sensor is determined by its kinetics. Additionally, the optical signal is detected in setups that determine aspects such as spatial resolution or imaging speed. In the following sections we discuss how these hallmarks of fluorescence sensors affect their performance.

87Chemical imaging with many SWCNT‐based sensors at one time is a highly effective strategy to gain chemical information from a sample with outstanding spatial and temporal resolution. To understand how the collective image of such an array of sensors reflects the concentration of an analyte, Meyer et al. used stochastic Monte Carlo simulations to study the kinetic requirements for spatiotemporal chemical imaging with nanoscale sensors such as SWCNTs.

88The subject of the simulation was a nanosensor array being exposed to a changing concentration gradient of an analyte. To predict the image one gains from many sensors, single sensor responses were first simulated. In a typical scenario, the time‐dependent concentration/diffusion profile of a dynamic process, such as release of signaling molecules (e.g. neurotransmitters) from cells and the stochastic binding site state of the sensors for certain rate constants, was simulated to calculate the expected fluorescence change of single sensors (Figure 20 a,b). The overall image was then calculated considering the individual fluorescence emission point spread functions and technical considerations, such as the frame rate of the camera. This simulation allows the prediction of how the rate constants of a sensor (k on, k off) and other factors affect the spatiotemporal resolution, for example, to resolve fast concentration changes such as neurotransmitter release from cells. It can serve as guiding principle for the chemical design but also for the interpretation of signals from a given biological problem. Phase diagrams (Figure 20 c) indicated that the senors need a surprisingly low affinity (K d=100 μm) to resolve fast processes.

89So far, optical biosensors based on SWCNTs have mainly been fabricated from mixtures of multiple chiralities. This leads to a large spectral overlap in the fluorescence emission, which complicates multiplexing and reduces sensitivity.

90Advances in the separation and functionalization of SWCNTs have recently enabled the sole use of SWCNTs emitting below 1000 nm together with low‐cost silicon‐based detectors for the development of ratiometric sensors, in which two or more distinct NIR signals are detected simultaneously. Here, one SWCNT chirality is typically not responsive to the analyte and acts as a reference. Besides the fact that single chirality SWCNTs lead to six‐ to tenfold higher fluorescence intensities compared to multichirality SWCNT mixtures at the same concentration, ratiometric approaches are more stable to external noise. Despite the clear advantages, the implementation was not possible for a long time because of difficulties in gaining chirality‐pure SWCNTs.

91Although progress was made in the synthesis of chirality‐enriched SWCNT samples, only a few chirality‐enriched samples are commercially available. Different separation and purification methods have been developed. They range from density gradient centrifugation, gel chromatography, ion‐exchange chromatography, and aqueous two‐phase extraction to wrappings of special macromolecules that preferably solubilize certain chiralities. What they all have in common is that the resulting pure chiralities are solubilized in certain polymers or surfactants. However, for a biosensing application it is necessary to tailor the surface chemistry. Therefore, straightforward processes are required that yield chirality‐pure SWCNTs with tunable functionalization.

92The first ratiometric SWCNT‐based sensor was demonstrated by Giraldo et al., who integrated into the leaves of plants two sensors: one for the detection of H2O2 and one for NO. The sensor system was based on the ratio of the distinct emission bands of two chirality species (Figure 21 a). Whereas the (GT)15‐(7,6)‐SWCNT‐based sensor was quenched by 20 % within 10 min in the presence of 100 μm H2O2, the reference sensor remained mostly invariant to the analyte. The overall sensor response was similar to in vitro tests.

93Most recently, Nißler et al. combined the isolation of specific SWCNTs and their subsequent functionalization for the detection of neurotransmitters and other small molecules. They used aqueous two‐phase extraction to obtain chirality‐pure (6,5)‐, (7,5)‐, (9,4)‐, and (7,6)‐SWCNTs and applied a surface exchange through dialysis to remove the DOC wrapping of the SWCNTs and replace it with specific ssDNA sequences or aptamers. This approach enabled the fabrication of ratiometric sensors, for example, for the detection of dopamine and H2O2 (Figure 22 b,c). For these sensors, SWCNTs were functionalized either with a (GT)40 sequence or, in the case of H2O2 detection, with an aptamer (hemin binding aptamer, HeApta) that binds the protoporphyrin hemin and catalyzes the decomposition of H2O2. PEG‐PL‐functionalized SWCNTs of another chirality served as an invariant reference. These examples show the potential of ratiometric or multiplexed sensing and given the large wavelength range of SWCNT fluorescence, there are plenty of opportunities for advanced sensing schemes.

94Remote imaging—the spatial separation of molecular sensors and detectors—is particularly beneficial to observe biochemical processes non‐invasively, for example, in biomanufacturing or in vivo. As only a small portion of the emitted light is captured by a camera within a certain distance to the sensors, remote detection requires either bright fluorophores or cameras with a high quantum yield in the spectral window of the fluorophore emission.

95An example for remote imaging are NIR fluorescent nanosheets, which were remotely detected in a customized portable setup. Due to the high brightness and emission closer to the visible range (910 nm), a low‐cost CMOS camera (Complementary Metal Oxide Semiconductor, Si based), which typically has a low quantum efficiency of around 5 % in the NIR region, could be used here.

96Additionally, this setup was modified for SWCNT detection by using an NIR‐sensitive InGaAs camera, a white‐light LED, and corresponding filters for wavelength‐specific excitation and detection of emission (Figure 22 a). Up to now, remote imaging of fluorescent SWCNT‐based sensors has been used for the identification of bacteria and monitoring of plant health, for which resolutions in the (sub‐)millimeter range were achieved with standoff distances of up to 1 m.

97Apart from exploiting the decreasing sensitivity of Si‐detectors in the NIR, efforts have been made to promote NIR‐fluorescent transducers using straightforward designs for inexpensive NIR fluorimeters. Instead of using high‐cost InGaAs photodiode arrays, single InGaAs diodes were combined with a motorized stage controlled with an open source programming language. The robustness of these devices outside of the laboratory was demonstrated in a high‐throughput format with field‐side measurements of soil samples. Future developments might also show remote detection using pure (6,4)‐SWCNTs with Si‐based cameras and also versatile use as smart surfaces, for example, for monitoring contamination with pathogens, such as bacteria, on medical surfaces or even implants.

98Fluorophore brightness is not the only performance‐related factor when measurements in live biological tissues are carried out. Fluorophore stability as well as wavelength‐dependent tissue scattering, absorption, and the background originating from autofluoresence must also be considered. From this point of view, SWCNTs have excellent photophysical properties.

99However, imaging at the single molecule level is always accompanied by considerations to achieve a high signal to noise ratio. Long‐term single SWCNT imaging has, in particular, become established for (6,5)‐SWCNTs in living cells and brain tissue.

100Although the second‐order excitonic transition, E22 (S22), is typically used for excitation, alternative excitation pathways such as the K‐momentum exciton‐phonon sideband (KSB) excitation and up‐conversion excitation have been successfully demonstrated, but excitation efficiency, photostability, as well as absorption and scattering of molecules in tissue differ depending on the excitation wavelength. Danné et al. evaluated the different excitation options (Figure 22 b) for optimal single (6,5)‐SWCNT imaging. Here, E22 excitation of PEG‐PL SWCNTs was found to be four times more efficient than KSB excitation and an order of magnitude more efficient than 1064 nm up‐conversion excitation, while the signal to noise ratio was more than five times higher for KSB and up‐conversion excitation. However, the excitation at the E22 transition is not ideal due to limited tissue penetration depth and autofluorescence. In addition, simulations to quantify the impact of tissue absorption showed that a higher temperature rise of the tissue induced by up‐conversion excitation might be an issue, thus suggesting that KSB excitation is the best choice when considering all the factors. Nevertheless, it still requires relatively high excitation doses in the kW cm−2 regime.

101Recently, single SWCNT imaging was performed in brain tissue in vivo by using ultralow excitation doses of 0.1 kW cm−2 (Figure 22 c). To achieve this, sp3 defects were introduced into (6,5)‐SWCNTs, which lead to a fluorescence emission at E11 * (1160 nm) when exciting at the first‐order excitonic transition E11 (985 nm). This approach is beneficial as a result of excitation in the NIR window, but also because of the increased brightness by channeling free excitons to defect sites and subsequent E11 * emission.

102Another way to both excite and detect in the NIR region is the use of multiphoton microscopy. This technique relies on the nonlinear excitation of at least two photons and is optimal for in vivo tissue applications when using NIR radiation. Del Bonis‐O′Donnell et al. demonstrated two‐photon 1560 nm excitation of dopamine‐sensitive SWCNT‐based sensors, which showed only 4 % scattering (one‐photon excitation: 42 % scattering). However, the frame rate is significantly reduced for single sensor imaging compared to wide‐field one‐photon excitation.

103Hyperspectral microscopy (i. e. simultaneous imaging at different wavelengths) provides another approach to exploit the spectral variety for multiplexed SWCNT imaging. Roxbury et al. resolved up to 17 distinct chiralities with single nanotube spatial resolution. In contrast to organic fluorophores, more chiralities can be imaged simultaneously in a certain emission window due to the narrow emission bands of SWCNTs. The hyperspectral imaging is based on the use of a volume Bragg grating (VBG) placed between the emission port of an inverted fluorescence microscope and the NIR camera. The VBG filters one specific emission wavelength depending on optical properties such as the incident angle θ and the grating period Λ (Figure 23 a).

104The spatial imaging of different chiralities is then achieved by measuring a continuous stack of 152 images within a time frame of 20 s and 10 min depending on the signal intensity and, thus, the integration time. By using this approach 12 different chiralities from DOC‐SWCNTs in live human cervical cancer cells could be detected (Figure 23 b). Additionally, individual SWCNTs adsorbed on a surface in live mammalian cells, murine tissues ex vivo, and zebrafish endothelium in vivo were imaged.

105Another challenge of typical wide‐field microscopy setups is the poor z‐resolution. To improve this limitation, a NIR spinning‐disc confocal laser microscope with an increased resolution and imaging contrast was demonstrated by Zubkovs et al. (Figure 23 c). The custom‐built microscope was based on a spinning‐disc module integrated between a cooled InGaAs camera and the microscope body, which rejects out of focus light.

106To achieve a maximized photon intensity, the lenses in the spinning disc unit were optimized for the NIR region. The authors showed in different biological applications the advantages of an improved lateral/axial resolution of 0.5±0.1 μm /0.6±0.1 μm (enhancement of 17 % /45 %) compared to the wide‐field configuration, reaching from single‐particle tracking over the spatial distribution of nanoparticles within an organelle to the optical in situ monitoring of glucose by GOx‐SWCNT‐based sensors embedded within an agarose gel. Overall, this approach showcases the opportunities for in vitro and in vivo imaging and sensing with improved spatiotemporal resolution.

107SWCNTs have shown their large potential as versatile building block for biosensors. The last few years have provided fundamental mechanistic insights, novel recations, and novel applications such as sensing in plants or primary cells. The advent of covalent quantum defect chemistry as well as the broader availability of monochiral samples will further advance typical figures of merit such as selectivity and robustness in demanding biochemical environments. One of the key challenges remains a basic understanding of molecular recognition and also signal transduction in the organic phase around these materials. Advances in this space can directly translate into superior selectivities. Additionally, high‐throughput or screening approaches will increase the speed of chemical discoveries. This is particularly interesting for applications in complex biological environments, where interactions become too complex to be predicted. Several milestones remain to be achieved before transfer into commercially available products. In vivo applications require long‐term evaluation of each sensor's toxicology and stability profile. Here, new covalent quantum defect functionalization strategies as well as the standardization of protocols and the purity of SWCNTs offer promising opportunities. Additionally, higher degrees of multiplexing and the adaptation of conventional cameras and readout systems will increase the application potential of SWCNT‐based sensors. In summary, SWCNT‐based biosensors offer a rich playground for chemistry and related disciplines and promise further advances and breakthroughs in the near future.

108Table 1, 2, 3, 4, 5, 6, 7, 8, 9

109The authors declare no conflict of interest.

110Julia Ackermann received her M.Sc. in nanoengineering with a focus on nano(opto)electronics at the university of Duisburg‐Essen (Germany). Since 2020 she has been persuing her Ph.D. in the technology division of biomedical nanosensors of the Fraunhofer Institute for Microelectronic Circuits and Systems in Duisburg (Germany). Her work focuses on improving the overall selectivity of carbon nanotube based sensors by pattern recognition.

111Justus Metternich received his Bachelor's degree in biotechnology from the University of Applied Sciences Darmstadt. After a short stay at the Centro de Investigaciones Biológicas Margarita Salas in 2018, he continued with his Master's studies in chemistry at Uppsala University. Since November 2020, he has been part of the Attract group of Sebastian Kruss and pursuing his Ph.D. at Ruhr‐University Bochum. His research focuses on the design of fluorescent carbon nanotube functionalizations for the detection of pathogens.

112Svenja Herbertz received her Ph.D. in Physics at the Solid‐State Physics Laboratory at the Heinrich‐Heine‐University in Düsseldorf (Germany) in 2019. During her years of study in Medical Physics she gained experience in the field of optical spectroscopy and use of semiconductor quantum dots for fluorescence labeling in biomedical applications. She is currently a researcher at the Fraunhofer Institute for Microelectronic Circuits and Systems in Duisburg (Germany) in the technology division of biomedical nanosensors.

113Sebastian Kruss received his Ph.D. in biophysical chemistry at Heidelberg University and the Max Planck Institute for Intelligent Systems (with Prof. Joachim Spatz). He then moved to the Massachusetts Institute of Technology (with Prof. Michael Strano), where he worked on carbon nanomaterials. After heading an independent research group at Göttingen University (2015–2020), he became professor of physical chemistry at Ruhr‐Universität Bochum and Attract group leader at Fraunhofer IMS. His research focuses on novel materials, fluorescence spectroscopy and microscopy, biosensors, and cell biophysics.