Nucleic acid sensors in vivo: challenges and opportunities

Nucleic acid biosensors, integrated with functional nucleic acids, have drawn extensive attention for diverse biomedical applications. In particular, many nucleic acid probes have been developed for biosensing and imaging of tumor‐associated species, including pH, small molecules, RNAs, metal ions, proteins, and cells. Despite the progress made, these sensors are mainly performed at the level of test tube assays or live cells for disease diagnosis and pathological studies. In recent years, increasing research works are pushing the limit of nucleic acid sensors in preclinical applications and imaging‐guided therapy. However, the applicability of current nucleic acid sensors in vivo is largely hindered by the complexity of living animals. Herein, we review recent advances in the design and applications of nucleic acid biosensors in vivo, in which the key factors for the fabrication of eligible sensors are highlighted. Additionally, given the inherent shortcomings of nucleic acid, we will also describe the challenges of the current in vivo nucleic acid biosensors and some new strategies that may significantly accelerate the development of biosensors.

advantage of the development of molecular engineering and nanomaterials, as well as an imaging technique, the applicability of NA sensors is pushed to the limit of living animals in recent years. Despite the progress have been made, current NA sensors are still unable to fulfill the requirements for in vivo imaging.
Although the performance of NA sensors at in vivo level can greatly facilitate their applications in the clinic and even for personalized medicine, the environment of the living system is more complex than that in a simple buffer or live cell, which poses a significant challenge for the construction of in vivo NA biosensors. A conventional NA biosensor is typically composed of two components including a target recognition group and signal reporter; 18 however, several challenges should be overcome before its application in vivo. First, the fluorescence reporters in the visible region, which is commonly used for NA sensors, suffer from insufficient tissue penetration depth and severe sample autofluorescence interference, resulting in low-quality of imaging performance in the in vivo environment. Second, the negatively charged NA has been demonstrated to possess a low cell membrane or tissue penetrability, therefore, appropriate carriers for the delivery of NA sensors are required. Third, the reactive NA sensors may be activated by the targets during their delivery to destinations, which causes obvious false positive signals and extremely low spatial resolution. Finally, the degradation of nucleases in the blood and the metabolism of NAs in the liver and kidneys make the half-life of unmodified NA biosensors in the living environment extremely short, which is not conducive to their sensing performance.
To address these issues, researchers have developed some smart and high-performance biosensors for target imaging in vivo, specifically, in tumor microenvironments (TME) 19,20 or tumors 21,22,23 and even in subcellular organelles of cancer cells. 24,25,26 Herein, to further accelerate the development of NA sensors and push their limit to preclinical applications and imaging-guided therapy, we reviewed the typical research works that achieve in vivo applications (Table 1), with the light shed on the strategies that raised by these works to overcome the challenges from complex living systems. Additionally, current difficulties and potential solutions were discussed to offer some possible improvements.

IN VIVO NUCLEIC ACID BIOSENSORS
To achieve superior performance in vivo imaging, researchers have integrated many advanced techniques into the fabrication of NA sensors. Toward various targets in vivo including pH, metal ions, small molecules, RNA, proteins and cells, different functional NAs with the rational design were chosen as recognition moieties. To increase the intensity of signal output, the fluorescent reporter with longer emission wavelength, as well as magnetic resonance (MRI), positron emission tomography (PET), and photoacoustic (PA) signal were introduced to increase the tissue penetration. Also, some DNA-reactionbased signal amplification approaches were employed to make the signal response to endogenous biomarkers of low abundance strong enough to be collected by imaging systems. In addition, some smart molecular switches are involved in the design of biosensors to improve the accuracy and spatial discrimination in target imaging. Therefore, this section will discuss the current work of typical NA biosensors in vivo from three parts: recognition groups, signal readout methods, and molecular switches, respectively.

Recognition elements
In addition to the natural property of NA hybridization for genetic target recognition, functional NAs obtained from in vitro selection, particularly aptamers and DNAzyme, have largely enriched the toolbox for non-genetic target detection. In this subsection, according to the different recognition groups, we will introduce in vivo NA biosensors according to their recognition moieties.

Aptamer
Aptamers, sometimes called chemical antibodies, are antibody-like molecules in that they function primarily in molecular recognition, having been comprehensively introduced in some other reviews. 58,59,60 In recent years, aptamers are increasingly employed as recognition groups against a wide range of targets, from small molecules to proteins, and even living cells and tissues. Among them, the cell-specific aptamers that screened from Cell-SELEX exhibit high binding capacity and superior specificity to cell membrane proteins, which provides an excellent candidate for in vivo tumor imaging and therapy. 61,62,63 Recently, Ye's group 27 has developed a multidrugresistant Hepatocellular carcinoma (HCC) cell-specific aptamer and realized tumor-targeted fluorescence imaging in vivo. In this work, the screened aptamer had a comparative binding affinity (Kd = 65.39 ± 21.85 nM) and could specifically accumulate in the drug-resistant tumors, which indicated its potential for precise diagnosis of multidrug-resistant HCC. However, the commonly used single-stranded aptamer can be easily degraded by nucleases, limiting theirs in vivo applications. To address this issue, Tan's group 64 proposed cyclic divalent NA aptamers that have largely improved resistance to nuclease degradation and target binding affinity ( Figure 1A-i). This work provided a simple but efficient strategy to implement NA aptamer-based sensors for in vivo tumor images. In addition, by cyclizing aptamers to different targets, a circular bifunctional aptamer consisting of Tau aptamer and transferrin receptor (TfR) aptamer was designed. 28 This bifunctional aptamer not only displayed enhanced plasma stability but also had an efficient blood-brain barrier (BBB) penetrating ability, which achieves the disruption of tauopathy and improves memory deficits in vivo.
In addition to tumor, ATP, a key metabolite in tumor progression, is also monitored to understand its physiological and pathological role in vivo. 65,66,67 ATP-aptamer, which undergoes a structural transformation when bound to ATP, has been widely used as a recognition element to achieve accurate detection and imaging of ATP. Typically, Li's group 19 has reported a TME -driven DNA nanomachine for specific ATP imaging in mice. The nanomachine is based on a structure-switching aptamer that can specifically bind to ATP, thus allowing a fluorescence response to extracellular ATP in the TME. Moreover, by combining ATP-aptamer with toehold-mediated strand exchange, they designed a DNA molecular circuit with ATP and miRNA as inputs and demonstrated that the circuit is capable of efficient "AND" logical computation in mice. 34

DNAzyme
As a functional NA capable of catalyzing substrate reactions in the presence of cofactors, DNAzyme is increasingly becoming a powerful tool in the detection and imaging of metal ions. Metal ions play an important role in physiological processes by assisting the function of many proteases, regulating the interaction of biomolecules and stabilizing the conformations of biomolecules. 68,69 Besides, abnormal distribution and expression levels of metal ions are often closely associated with a variety of diseases such as osmolarity imbalance, Alzheimer's disease and even cancer. 70,71,72 Researchers have taken advantage of the excellent specificity of DNAzyme for metal ions to construct many biosensors with excellent imaging performance, achieving monitoring of metal ions at the living cell level.
Currently, some DNAzyme-based biosensors enable metal ion imaging at the in vivo level. Lu's group 35 has developed a near-infrared (NIR) DNAzyme nanoprobe, achieving the real-time monitoring of Zn 2+ within early embryos and larvae of zebrafish ( Figure 1A-ii). Moreover, to avoid the lack of tissue penetration depth inherent to light-activated probes, their group has recently reported a novel DNAzyme-based biosensor, which utilizes highintensity focused ultrasound (HIFU) to realize the spatiotemporal imaging of metal ions in mice at the targeted region. 36 To achieve specific detection of metal ions by distinguishing different cells, Li's group 21 has developed an enzyme-activated DNAzyme biosensor by introducing an abasic site that can be cleaved by cancer-specific enzymes, thus realizing the precise imaging of metal ions in tumor-bearing mice.

Others
In addition to the in vivo NA biosensors mentioned above, which use aptamer and DNAzyme as recognition groups, there are also some biosensors designed by the hybridization of DNA/RNA with molecular beacons, a functional NA with a hairpin-type structure. These molecular beacon-based biosensors can specifically identify with NA sequences and trigger a subsequent series of strand hybridization reactions. Since they are often involved in the design of signal amplification probes such as hybridization chain reaction (HCR) and catalyzed hairpin assembly (CHA), which use continuous crosstalk hybridization between metastable hairpin DNAs to achieve effective signal amplification, we will describe them in the signal amplification section. At the same time, the specific interaction of endogenous disease-associated enzymes with their substrates is often used as a recognition modality for biosensors. Telomerase (TE), a typical biomarker for cancer diagnosis, can recognize its specific sequence and extend this NA sequence. 22,23 Therefore, there has been a lot of work in recent years to involve the specific binding of telomerase to specific sequences as a recognition modality in the design of biosensors for telomerase activity detection. Yang's group 39 has constructed a telomerase-activated MRI probe, achieving the monitoring of telomerase activity in vivo. The probe displaces down the paramagnetic complexlabeled NA strand through recognition and extension of the primer by telomerase, thus enabling a distancedependent shift of the MRI signal from off to on. In addition, apurinic/apyrimidinic endonuclease 1 (APE1), a DNA repair enzyme that is often overexpressed in cancer cells, can specifically cleave the abasic site of NA substrates. 73,74 Therefore, APE1 is receiving more and more attention in the field of early diagnosis and imaging of diseases. Li's group 24,25,26 has presented an NIR light-activated nanoprobe that can track the subcellular dynamics of APE1 during the photodynamic therapy. Moreover, by designing the TE-primer and APE1-cleavable site on the same double-stranded DNA probe, their group has developed a cooperatively activatable NA biosensor to detect the correlated activity of TE and APE1 in vivo ( Figure 1A-iii). 45 In addition, Min's group 46 has performed in vivo imaging of protein kinase A (PKA) activity through the disassembly of the DNA/peptide complex in which PKA can cause electrostatic inversion of its substrate peptide during phosphorylation to release the initiator chain.

Enhancement of signal collection
When NA biosensors are applied to in vivo imaging, the effective output of the target identification signal is a challenge. The challenge basically stems from two aspects, on the one hand, in vivo imaging requires deep tissue penetration, in which the commonly used fluorescent signal cannot meet the demand. On the other hand, many biomarkers are expressed in low abundance in vivo which is not enough to produce an obvious signal change. To address these issues in signal collection, current in vivo NA biosensors mainly focus on increasing tissue penetration or amplifying signal intensity to achieve effective acquisition of signals from in vivo targets. In this subsection, we will introduce the current signal collection methods for in vivo sensing in terms of both deepening tissue penetration depth methods and signal amplification strategies.

Increasing tissue penetration
Although NA biosensors have good specificity and sensitivity with fluorescence as the signal output mode. The limited tissue penetration depth and strong autofluorescence interference of fluorescence imaging greatly limit the application of NA biosensors in vivo. Combining deep tissue penetrating imaging modalities such as MRI, PET imaging, PA imaging, etc. with NA biosensors is a simple and effective strategy. By combining in vitro Cell-SELEX and in vivo screening technology, Chen's group 47 has selected a HER2-targeting DNA aptamer and realized rapid and specific HER2 in vivo imaging. As shown in Figure 1B-i, this work enables PET imaging of HER2 by modifying the NA sequence with. 68,69 F, thus avoiding the problems of severe background and low penetration depth of fluorescence imaging. The 1-3 mm tissue penetration depth of PET imaging allows for better real-time imaging of tumor sites in live mice. Lu's group 50 has demonstrated the first aptamer-based (PA) biosensor for activatable PA imaging of thrombin in living mice. By modifying DNA strands with NIR fluorophores/quencher pair, they have achieved an effective combination of the activatable properties of NA probes with high resolution at a deep penetration depth of PA imaging. In addition to PET and PA imaging, Wang's group 39 developed a telomerase-activated magnetic resonance (MR) imaging probe for telomerase detection in vivo. Paramagnetic Gd-DOTA ((Gd(III) 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraazacyclododecane-1,4,7,10-tet-raacetic acid) complexes was conjugated to the oligonucleotides, which hybridized with anchor DNA. And one other segment of anchor DNA is complementary to telomerase primer strand. Since the anchor DNA is covalently modified on the surface of superparamagnetic ferroferric oxide nanoparticles (SPFONs), the close distance between Gd-DOTA and SPFONs made the T1 MRI signal of Gd-DOTA quenched. In the presence of telomerase, the extension of telomerase primer strand would displace Gd-DOTA conjugated DNA from the surface of SPFONs, inducing an enhanced T1 MRI signal for in vivo detection of the telomerase activity. Besides, Wang's group 51 developed a multiple-armed DNA tetrahedral nanoprobe modified with radioactive isotope 99m Tc and NIR dyes Dylight 755, achieving tumor-targeting single-photon emission computed tomography (SPECT) and NIR fluorescence imaging in tumor-bearing mice.

Signal amplification
Since NA biomarkers such as mRNA and miRNA are in low abundance in vivo, with the development of DNA nanotechnology in recent years, many NA signal amplification techniques such as HCR, CHA and enzyme-based amplification strategies have been developed. 29,75 These NA signal amplification methods enable sensitive detection and imaging of a wide range of disease-associated NAs at the living cell level. To further advance the application of these signal amplification methods in the field of early disease diagnosis, researchers have attempted to image target NAs with amplified signals in vivo. Li's group 31 reported a molecular beacon-based enzymatic fluorescence amplification that achieved in vivo imaging of inflammation-associated mRNA in a mouse paw model with acute inflammation. In this work, the loop part of the molecular beacon probe is modified with two apurinic/apyrimidinic (AP) sites, so that after hybridization with the target mRNA, the cleavage by APE1 enzyme enables the recycling of the target mRNA and thus effective signal amplification. Moreover, recently they further developed an enzymatically triggerable signal amplification for spatially selective sensing of ATP in inflammatory cells. 33 In the presence of APE1, the structure-switching activity of anti-ATP aptamer probe (bearing with AP site) can be restored, inducing the release of the complementary DNA strand. Subsequently, the complementary DNA strand can be used as the initiator to hybridize with hairpin DNA modified with AP site, thereby enabling APE1-mediated signal amplification. Since APE1 is only present in the cytoplasm of inflammatory cells but not normal cells, this strategy realized inflammatory cell-selective amplified ATP imaging in vitro and in vivo.
In addition to endogenous enzyme-based signal amplification methods, NA signal amplification strategies without protease involvement such as HCR and CHA also have great potential in the field of in vivo imaging. Wang's group 53 developed proteinase-free DNA replication machinery by combining HCR with DNAzyme, and analyzed tumor-related miRNA in living mice by intratumoral injection of DNA machinery. To achieve high-fidelity imaging of microRNAs in vivo, they have further designed an orthogonally controlled catalytic DNA (CCD) circuit which utilizes multiple guaranteed molecular recognition ( Figure 1B-ii). 54 Through pre-mixed with lipo3000, CCD probe exhibited high robustness for in vivo high-fidelity miR-21 imaging. Furthermore, they recently constructed an autocatalytic hybridization assembly (AHA) circuit which combined cascade hybridization reaction (CHR) with catalytic DNA assembly (CDA). 55 The circuit can generate exponentially amplified signals responding to target miRNA, thus can achieve sensitive and selective imaging of miRNA in mice.
In order to avoid unsatisfactory detection performance of multi-component biosensors in in vivo environment due to free diffusion of each functional element, using DNA nanomaterials such as Y-shaped DNA structure and DNA nanowires to integrate each component spatially and stoichiometrically has been a smart and effective strategy. Jiang's group 37 has developed a tripartite DNA probe integrating aptamer with target specificity and components of CHA. This work realized mRNA fluorescence imaging in living animals and provided an improved strategy for early-stage clinical diagnosis. This group also applied the strategy to an HCR and realized high-contrast imaging of miR-21 in living mice. 38

Molecular switches
In a complex in vivo environment, the target molecules usually not only exist in the destination location but also exist in other tissues at some levels. Thus, the biosensors are prone to be activated by targets at non-interested locations during in vivo transport, resulting in false positive signals and low spatial resolution. To meet this challenge, various molecular switches that can be triggered by external or endogenous stimulations were introduced into the design of NA biosensors in order to improve the accuracy as well as the spatiotemporal resolution of target detection and imaging.
As a non-invasive stimulation, light has been widely used in recent years to activate NA biosensors. However, the commonly used photocleavable (PC) groups, that require UV light for activation, are not applicable in vivo due to the extremely low tissue penetration of UV light. 76 A fantastic approach is converting the trigger light of good tissue penetration (e.g., NIR) to UV light locally. Li's group 57 functionalized UCNPs with a designed DNA beacon that was modified with a PC group. Upon NIR light irradiation, UCNPs can convert NIR light to UV light to decage the PC protection group, resulting in target miRNA hybridizing with a fluorophore-labeled strand, thus achieving photo-activatable imaging of miR-NAs in living mice with high spatiotemporal resolution.
They also introduced the PC group to other ATP aptamer/imotif/AP-site-based biosensor, achieving ATP imaging/pH sensing/APE1 monitoring in tumor-bearing mice with spatiotemporal control (Figure 1C-i). 25,[40][41] Though non-invasive lights have been demonstrated to control the activation of NA sensors in vivo, this method requires the introduction of nanomaterials of potential toxicity. Alternatively, the endogenous stimulus may also be used to activate the NA sensors. Cleavage of the substrate chain by proteases can often cause changes in the structure of the probe. Moreover, overexpression of many endogenous proteases is often closely associated with the disease. Therefore, the specific recognition and interaction of these proteases with substrate chains as molecular switches of biosensors can further improve the accuracy of disease diagnosis at the in vivo level. Li's group 42 developed a conditional DNA aptamer sensor using the hydrolysis of substrate peptides by cathepsin B (CaB) as a switch for probe activation. They achieved tumor-specific protease-controlled ATP imaging in vivo. In addition to CaB, they also constructed an enzyme-regulated amplification strategy using the AP site-specific cleavage of APE1 as a molecular switch coupled with CHA signal amplification ( Figure 1C-ii). 43 In this work, tumor-cell-specific miRNA amplified imaging in vivo with enhanced spatial accuracy was achieved. Wang's group 44 also used a similar strategy to achieve precise and amplified imaging of miR-21 in mice.
In addition to the aforementioned endogenous stimulus, several other molecular switches that rely on the TME are currently involved in the construction of in vivo biosensors. Li's group 19 reported a TME -driven DNA nanomachine, utilizing the pH (low) insertion peptide (pHLIP) module to achieve specific anchoring on the surface of the tumor cell membrane, thus allowing a fluorescence response to extracellular ATP in the TME. They also used the breaking of disulfide bonds by glutathione (GSH) as the molecular switch to activate ATP aptamer-based nanoprobe and demonstrated correlated imaging of ATP and GSH in living mice. 24 To improve the specificity, Tan's group 32 developed a hypoxia-activated aptamer, which would show great potential in the precise diagnosis of tumors with a unique hypoxic microenvironment.

Stability
Despite the progresses NA biosensors have made in the field of in vivo imaging, there are still many challenges for NA-based biosensors to further promote their performance in vivo due to the inherent defects of NAs. The first defect is the poor stability of NAs in complex biological environments. The degradation by nucleases in the blood and the metabolism by the liver, kidneys, and other organs make the half-life of unmodified NA biosensors extremely short in the living system, hindering the accuracy and the continuous observation of targets. Though the NA nanoprobes which used nanomaterials as carriers were reported to prevent the sensor from degradation to a certain extent, the adhesion of serum proteins on the nanoparticles may also abolish the functions of the sensors. Also, the metabolism of the nanoparticles may cause hepatotoxicity, which reduces the biocompatibility of NA sensors. The emerging DNA nanostructures, such as DNA nanotube, 77,78 DNA polyhedron, 79,80 and DNA origami 81,82 can perfectly address these issues, but the high cost of these DNA nanostructures hinders their applicability in animals. And there is a lack of effective and smart ways of unloading cargo when using DNA nanostructures or even DNA origamis for sensor cargo delivery, which can also greatly improve the use of NA sensors for in vivo imaging and detection.
In recent years, the oligonucleotides consist of unnatural NAs, such as locked nucleic acid (LNA), 48,49 peptide nucleic acid (PNA), 83 2'-fluoroarabinonucleic acids (FANA), 84 (3'−2') a-l-threose nucleic acid (TNA), 52 and xenonucleic acids (XNA) 85 , have been demonstrated to possess superior stability in the living system because of their resistance to the degradation of natural nucleases. However, straightforward replacing the NAs in current functional NA may entirely abolish their recognition capability because the sensors compose of unnatural NAs are unable to maintain the original second or tertiary structures. To meet this challenge, Zhu's group 86 has developed a 'mirror-image selection' strategy, which could utilize mirror-image DNA polymerase to select L-DNA aptamers. The L-DNA aptamer screened by this method not only exhibits good biostability but also possesses a comparable recognition capability to the current DNA aptamer. It is believed that the L-DNA will bring new opportunities for NA biosensors in the field of disease diagnosis and imaging at the in vivo level.

Signal readout
Most of the current in vivo NA biosensors use fluorescence as signal outputs. Although fluorescence imaging has the advantage of high sensitivity and resolution, this imaging mode exhibits low tissue penetration depth and severe autofluorescence, resulting in poor-quality of imaging. This issue can be addressed by the integration of fluorescent dyes of long excitation wavelengths, such as NIR-II dyes, 87,75 into NA sensors. However, as a class of emerging dyes, no quencher molecules were developed to match the NIR-II dyes yet, which hinders the fabrication of turn-on sensors. Potentially, some broadspectrum quenchers made up of nanomaterials, such as gold nanoparticles (AuNPs), 88 graphene oxide (GOx), 89,90 and MnO 2 nanosheets, 91,92 can serve as the quenchers for NIR-II NA sensors. Furthermore, chemi-and bioluminescent sensors have advantages in their signal-to-noise ratio and response dynamics and do not require an excitation light source. 95 These advantages may make them valuable for generating novel NA sensors for in vivo imaging. However, building the connection between chemi-or bioluminescent signal modes and NA sensors is still a challenge. Though several attempts have been made to construct the chemi-and bioluminescent NA sensors, none of them achieved in vivo imaging. 93,94,95 Alternatively, some afterglow optical agents, such as the semiconducting polymer nanoparticles (SPNs), that do not require realtime excitation also possess the potential to fabricate the excitation-free NA sensors. 96,97

CONCLUSION
Applications of NA sensors in vivo would enable the development of preclinical research and imaging-guided therapy. In this minireview, current works on the applications of NA sensors in vivo were summarized, which focused on the key elements for the fabrication of the in vivo NA sensors. The grand challenges of this trend were also discussed, and a personal outlook on potential directions and opportunities was addressed. We believe the integration of advanced techniques in molecular engineering, nanomaterials and imaging technologies would break the bottlenecks that NA sensors have in vivo, offering powerful analytical tools for biomedical imaging.

C O N F L I C T O F I N T E R E S T S TAT E M E N T
The authors declare no conflict of interest.