Biosafety of inorganic nanomaterials for theranostic applications

Recent advances in inorganic nanomaterial-based theranostics enabled imaging-guided molecular targeting and drug delivery, and various combinations of theranostic systems. The term “theranostics” is defined as diagnosis processed with therapy simultaneously with a specific connection between therapy and diagnosis. The inorganic nanomaterials, representatively carbon, metal, ceramic, and semiconductor-based nanomaterials, exhibit their unique characteristics to be used in theranostic applications. However, the unveiled human biosafety of nanomaterials for clinical use has become a major concern. Therefore, in this review, we compiled recent research on in vitro and in vivo biosafety of inorganic nanomaterials in various theranostic applications, along with a discussion of how the particle formulation, size, surface functionalization, test species, and test condition affect biocompatibility. Furthermore, the progress and challenges of the development of biocompatible inorganic nanomaterials for theranostic applications were discussed. In conclusion, with appropriate precautions on the biosafe condition to be administered, inorganic nanomaterials can be proposed to have excellent potential in the future theranostic application.


Introduction
Theranostics was defined by Funkhouser in a press released by PharmaNetics to describe a new business model and has been increasingly referred to in the biomedical research since it started in 2002 [1,2]. The term "theranostics" is defined as diagnosis processed with therapy simultaneously with a specific connection between therapy and diagnosis [3]. For example, specific theranostic probes enable one or more clinically important functions such as early detection of tumors by pinpointing their location and condition, delivering drugs specifically to targeted cells, and visualizing the efficiency of drugs on the targeted cells [4]. As well as imaging-guided molecular targeting and drug delivery, various combinations of theranostic systems are being developed. Photoablation therapy such as photodynamic therapy (PDT) [5] and photothermal therapy (PTT) [6] is often combined with numerous molecular imaging systems including magnetic resonance imaging [7], positron emission tomography (PET) [8], single photon emission computed tomography (SPECT) [9], photoacoustic imaging (PAI) [10], and fluorescence imaging system [11], leading to the emergence of precision medicine [12][13][14]. To enhance the precision and efficiency of the theranostic system, material scientists focus on the development of novel theranostic probes by utilizing the extraordinary properties of nanomaterials [15].
Nanomaterial is intentionally produced in nanometersized materials and usually 1-100 nm scales but loosely refers to nanoscale above 100 nm if they exhibit the specific characteristics of nano-sized materials [16]. At the nanoscale, the general physicochemical properties of the material can be completely different from those of their bulk counterparts by various parameters such as size, formulation, specific surface area, and surface chemistry [17]. Especially, inorganic nanomaterials including carbon Moon Sung Kang and Mina Kwon contributed equally to this work. nanomaterials (CNMs), metal nanomaterials, ceramic nanomaterials, two-dimensional (2D) nanomaterials, and semiconductor nanomaterials exhibit their unique physicochemical properties [18][19][20]. Compared to their organic counterpart, inorganic nanomaterials feature relatively high chemical stability, inertness, ease of functionalization, and novel optical, electrical, and magnetic behaviors [17]. The high surface-volume ratio of nanomaterials makes them have exceptionally high drug loading capacity [21]. The diversity of surface functionalization and material conjugation endow them with multimodality [22]. The modulation of their size and formulation leads to tailored physicochemical and optical characteristics for imaging applications [23]. Moreover, remarkable phenomena emerge or are reinforced in nanomaterials, such as the photoacoustic (PA) effect [23], enhanced permeability and retention (EPR) effect [24], localized surface plasmon resonance (LSPR) effect [25], and photodynamic (PD)/photothermal (PT) effect, which promises great potential in theranostic application [26,27]. Especially, these characteristics have been intensively studied for targeted cancer therapy. The EPR-mediated selective tumor accumulation, tumor-selective drug delivery, and pharmacokinetic profiles of chemotherapeutics can be tailored by control of nanomaterial size, shape, targeting ligand, and tumor microenvironment-selective drug release system [28]. Furthermore, the use of inorganic nanomaterials as imaging agents for cancer diagnosis highly improved imaging sensitives and accuracy. For example, magnetic nanomaterials have been highlighted as their theranostic properties. They can be combined with fluorescent molecules or anti-tumor drugs and used as magnetic resonance imaging (MRI) agents [29][30][31].
However, the human biosafety of nanomaterials for clinical use has become a major concern. Unlike the bulk state, nanomaterials induce different chemical or mechanical reactions with the human body that can cause unexpected side effects [32]. Most of the inorganic nanomaterials exhibit acute and chronic cytotoxicity both dose-and time-dependent manners; therefore, the administration condition of inorganic nanomaterials should be carefully tailored [33,34]. The reduction in particle size that catalyzes the surface chemistry of nanoparticles alters zeta potential, dielectric constant, and surface charges [29]. Generally, the role of surface chemistry is getting severe at single particle and small aggregates, rather than large aggregates. The small-sized nanomaterials inside or beside the human cells often persist in the system for elongated times due to the incapability of phagocytosis by macrophages. These phenomena lead to the generation of unwilling by-products and immune reactions leading to severe health risks such as disruption of the cell membrane and organelle integrity, genotoxicity on the nucleus, and intracellular reactive oxygen species (ROS) production [35][36][37][38][39]. Especially, the generation of ROS is one of the most emphasized causes of nanomaterial toxicity.
Under the excessive ROS exposure induced by nanomaterials, the natural antioxidant system of cells can be impaired [40]. The high amount of ROS can result in damage to biomolecules and organelle structures by protein carbonylation, lipid peroxidation, deoxyribonucleic acid (DNA) breakage, and finally bring out necrosis or apoptosis of cells [41]. Accordingly, the clinical application of inorganic nanomaterials should be cautiously employed despite their theranostic advantages. Despite these issues, the majority of nanomaterial-based publications within decades have focused on the fabrication and development of nanomaterials, and fewer are focused on the biological impacts. While the toxicity of bulk materials is well known, it is still unveiled at what concentration or size they can start to exhibit unknown toxicological properties due to their nanoscopic dimensions [42]. Furthermore, direct comparison of nanomaterials is hurdled by unreliable results derived from differences in culture conditions and the inconsistency between cell types and animal species [43].
Therefore, we demonstrate that there is a need for sufficient consideration of the biosafety of inorganic nanomaterials. Without a clear establishment of the rationale for biosafety issues, the clinical application of inorganic nanomaterials would be too insecure work. Their remarkable theranostic potentials as next-generation nanotheranostics, such as multimodal property, tumor pinpointing, targeted drug delivery, photobiomodulation, and high-resolved imaging system, can be realized after a better understanding of toxicity mechanisms as well as their impact on human health. This review categorizes inorganic nanomaterials by class and aims to summarize in vitro and in vivo biosafety studies for each material (Fig. 1). To avoid providing only superficial information, this review decided not to encompass the extensive range of every inorganic nanomaterial. Instead, this review focus on the representative materials which are preliminarily used in theranostic studies, by categorizing them into four sections including carbon, metal, ceramic, 2D, and semiconductor-based inorganic nanomaterials. At the same time, we compare the biosafety assays of several studies and discuss the potential toxic mechanism of each inorganic nanomaterial. Afterward, we will discuss ways to improve the biosafety of inorganic nanomaterials by physicochemical modification and expect that they can be used as basic data for future research and clinical applications. We hope that the biosafety study of inorganic nanomaterials guides us to break new ground for the next generation nanotheranostics.

Biosafety assessments
To clarify the underlying mechanisms of nanomaterial toxicity, multifaceted toxicological aspects and their assessment methods should be considered (Fig. 2) [44]. Although in vitro assays provide partial information on the biosafety of materials, essential validation can be evaluated before in vivo application to various experimental conditions and species. Afterward, in vivo assays should be conducted to elucidate the biodistribution, clearance pathway, and several potential side effects. Various in vitro and in vivo assays have been developed to provide toxicity mechanisms regarding the biosafety of inorganic nanomaterials (Table 1). These  assays clarify the type of cell damage which can be derived from chemicals, physical damage, radiation, microorganism, and other varied factors [45].
The 3-(4,5-dimethylthiazol-2-yl)2,5-diphenyl tetrazolium bromide (MTT) assay is one of the most common assays for evaluating the metabolic activity of cells by measuring the activity of nicotinamide adenine dinucleotide phosphate (NADPH)-dependent cellular oxidoreductase enzymes. When the living cells are treated with yellow tetrazolium dye MTT, it is reduced by reductase in mitochondria to form a crystal called formazan [46]. The formazan is soluble in specific solvents such as acidified ethanol and dimethyl sulfoxide (DMSO) to produce a purple-colored solution. Subsequently, the absorbance of the solution is detected between 500 and 600 nm using a microplate reader. Except for MTT assay, other tetrazolium-salt-related colorimetric assays such as 2,3-bis-(2-methoxy-4-nitro-5-sulfophenyl)-2H-tetrazolium-5-carboxamide (XTT), 3-(4, 5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS), and water-soluble tetrazolium (WST) can be used to evaluate metabolic activity with their own advantages [47]. The Alamar blue assay kit is a resazurin-based solution that is utilized as a cell health indicator by using the reducing power of cells. Resazurin is a blue cell-permeable compound that is reduced to fluorescent resorufin after cell uptake.
Apoptosis is another common cytotoxic mechanism of cells by nanomaterials, which is a programmed cell death processed through blebbing, cell shrinkage, nuclear fragmentation, chromatin condensation, DNA fragmentation, and messenger ribonucleic acid (mRNA) decay. The cell fragments called apoptotic bodies are quickly engulfed by the phagocytes to prevent spilling out of the cell contents into surrounding cells [48]. Annexin V staining is a commonly used apoptosis assay kit. The phosphatidylserine of normal cell phospholipids locates in the inner membrane; however, the phosphatidylserine of apoptotic cells moves to the outer membrane. The fluorescein-combined Annexin V solution selectively binds to phosphatidylserine of the outer membrane that can be quantified by fluorometric assays. The TdT-mediated dUTP-biotin nick end labeling (TUNEL) assay assesses the enzyme terminal deoxynucleotide transferase (TdT), which binds to deoxynucleotides to the 3′-hydroxyl terminus of DNA breaks. The distribution of apoptotic cells in the cell and tissue levels can be observed by fluorescence microscopy, optical microscopy, and flow cytometry [49]. Meanwhile, intracellular ROS production is a common cytotoxic factor induced by inorganic nanomaterials [50]. Normally, cellular antioxidant enzymes and free-radical scavengers protect cells from oxidative stresses. However, when excessive ROS is produced by external nanomaterials, the oxidative stresses begin to inflict damage on cellular macromolecules, such as lipids, proteins, and nucleic acids, which finally lead to pathological conditions [51]. The dichlorofluorescein diacetate (DCFDA) assay is an intracellular ROS assay and free radical sensor, which is a typical oxidative stress indicator to directly evaluate cellular redox states. When DCFDA passively enters a cell, the two ester bonds are broken, leading to the production of H 2 DCF and accumulation by ROS in cells [52]. Consequently, the oxidized cells are represented as highly fluorescent dichlorofluorescein (DCF), which is proportional to the intracellular ROS level.
Cell membrane integrity is another criterion of cytotoxicity, which is derived from direct contact between inorganic nanomaterials and cells. The inorganic nanomaterials often induce physical stress to cell membranes, leading to the breakage of membranes and necrosis [53,54]. Using vital dyes is a conventional method to assess membrane integrity. Neutral red is a weak cationic dye that passively enters the cell membrane and lysosome at a normal state. When the lysosome and membrane are damaged, the internal dye is released outside and the absorbance at 540 nm is assessed. Trypan blue is an acidic dye that strongly combines with proteins. The internalized trypan blue is exocytosed outside by live cells but remains in the cytoplasm in the dead cells, which is observed by microscopy. To measure the leakage level of the cytoplasmic enzyme such as lactate dehydrogenase (LDH), the degree of membrane damage can be evaluated. After cell membrane damage, LDH releases outside of the cell and oxidizes lactate to generate nicotinamide adenine dinucleotide-H (NADH), then reacts with WST to generate a yellow color. Using a microplate reader, the intensity of yellow dye correlates directly with the number of cells with damaged membranes [55].
On the other hand, the number of live and dead cells can be visualized by several kinds of live/dead assays [56]. The calcein AM and ethidium homodimer-1 are the most widely used live and dead cell dye pairs. In live cells, non-fluorescent calcein acetoxymethyl (AM) is converted to green fluorescent calcein after hydrolysis of acetoxymethyl ester by intercellular esterase, while the ethidium homodimer selectively penetrates the dead cells and attaches to nucleic acids to emit red fluorescence. On a similar basis, fluorescein diacetate (FDA) and propidium iodide (PI) stainings are another live/dead assay method using the selective entrance of FDA and PI solution into the cells [57]. PI is membrane impermeant and therefore excluded from viable cells, however, binds to DNA and ribonucleic acid (RNA) of dead cells and exhibits red fluorescence, while FDA is a cell-permeant esterase substrate and their green fluorescence by the enzymatic activity of cells. Meanwhile, flow cytometry allows fast, relatively quantitative, multiparametric analysis of cell populations at a single cell level, using the laser beam that hits the single cell moving in a fluid system [58]. A large number of functional membrane proteins involved in cell interaction, adhesion, and metabolism can be measured on the same cells making it become a useful tool for identifying and sorting cells within complex cell populations [58].
The genotoxicity of inorganic nanomaterials should be also considered. Since the inorganic nanomaterials can cross the cell membrane and reach the cell nucleus, down-regulation of specific genes, alteration and breakage of DNA, and phenotypic alteration can originate that lead to severe defects in cellular behaviors [59,60]. The genotoxicity can be assessed by evaluation of DNA damage using the comet assay, also known as the single-cell gel electrophoresis (SCGE). It measures the DNA damage at a single cell level using electrophoresis using cell lysates. Normal DNA strands with large and organized structures are fixed on the gel, while the broken DNA fragments are released from the DNA strands and can migrate along the gel. One can easily find the fragmented DNA migration which seems to be similar to comets [61].
Immunohistochemical analysis is a laboratory method that uses antibodies to check for certain antigens (markers) at a tissue level or biopsy. The species are treated with a series of processes including fixation, washing, dehydration, cleaning, paraffin infiltration, embedding, and cutting. Although many kinds of staining have been developed, hematoxylin and eosin (H&E) staining and Masson's trichrome staining are commonly used for biosafety assays. The hematoxylin stains cell nuclei as blue and eosin stain the matrix and cytoplasm pink. They provide valuable information on the pattern, shape, and structure of cells in a tissue-level sample [62]. Meanwhile, Masson's trichrome is a multi-colored staining dye that stains keratin and muscle fiber as red, blue on collagen and bone, red on cytoplasm, and dark brown on cell nuclei [63]. It is often used for the diagnosis of muscular dystrophy, cardiac infarct, hepatic cirrhosis, and kidney pathologies [64,65].
Although gene/protein and biochemical marker evaluation are not directly related to biosafety evaluation, they are widely used to find the mechanism or assess the expression level of cytotoxic markers. Polymerase chain reaction (PCR) means enzymatic amplification of nucleic acids to assess the gene expression and has the advantage of being able to amplify a very small amount of a sample in a large amount. To overcome the impossibility of quantifying the amplified product, quantitative reverse transcription (qRT)-PCR for quantitatively detecting the amount of amplified nucleic acid has been used [66]. The level of gene expression also can be measured by microarray analysis, which is a method that uses microchips containing anchored arrays of short DNA elements for the massive assessment of gene expression. The microarray analysis is easy to use, does not require DNA sequencing, and allows the quantification of thousands of parallel genes from different samples [67]. Meanwhile, the western blot is often used in cytotoxicity research to separate and identify proteins. The unknown blended proteins are separated based on their molecular weight through gel electrophoresis; hence, the type and quantity of proteins are revealed [68]. The enzyme-linked immunosorbent assay (ELISA) is a method of quantifying the amount of antigen protein contained in a biological sample using an antigen-antibody reaction to analyze the amount of antigen by measuring the activity of an enzyme bound to an antibody. Numerous cytokines concerning cytotoxicity or inflammation can be measured by ELISA [69].
Except for the explained methods, many evaluation methods have been to support the exploration of cytotoxicity mechanisms. The microscopic observation using colony formation assays (clonogenic assay), cellular uptake assay using transition electron microscopy (TEM), and atomic emission spectroscopy (AES) help to provide morphological information at cell levels. The complete blood panel analysis and hemolysis assay are used to evaluate hemotoxicity. Moreover, physicoelectrical assessments such as cell-impedance measurement, multisizer assay, and magnetic susceptometry can be used. On the other hand, in vivo safety assessment standards such as immune reaction, irritation, sensitization, and pyrogenicity can be considered to ensure the clinical safety of inorganic nanomaterials. The following section focuses on the detailed biosafety of inorganic nanomaterial to suggest guidelines for further safe theranostic applications.

Biosafety of CNM-based nanotheranostics
The CNM exists as various allotropes including nanodiamond, fullerene, graphene (G) derivatives, CNT, carbon dot (CD), and carbon nanoparticle (CNP), which have extraordinary physicochemical characteristics. Especially, the unique optical and electrochemical properties make them a potential probe for theranostic application [81]. The diverse physicochemical characteristics of CNMs are derived from the variation in sp, sp 2 , and sp 3 hybridizations. The transition from sp 3 to sp 2 carbons induces different carbon structures and the ratio of sp/sp 2 /sp 3 hybridizations determines which allotrope they will become [82]. Especially, G and CNTs are attractive platforms for theranostic application because they facilitate cellular uptake and ease of surface functionalization with various biomolecules to improve drug delivery efficiency [83][84][85]. Owing to their unique optical properties, CNMs have also been explored as contrast agents for PAI and PA tomography (PAT) enabling high-resolution imaging of cells, tissue, and organs as well as metabolic or pathological changes [82]. However, their potential toxicity is a major concern, which can be derived from their extremely small sizes and large specific surface area [33]. Due to the diversified allotropes and shapes, each CNM exhibits a different mechanism of toxicity that depends on cell types [86].
G is one of the two-dimensional (2D) nanomaterials that is composed of monolayered sp 2 -bonded hexagonal carbon atoms [87]. G is obtained by kinds of methods such as mechanical exfoliation, liquid phase exfoliation, chemical vapor deposition, electric arc production, and thermal decomposition [81]. Because of their low production cost, large-scale production, and easy processing, it has been one of the most frequently used CNMs. GO contains rich active oxygen-containing functional groups, which are used as catalytic active centers for covalent and noncovalent bonding with desired molecules or drugs [88]. Reduced GO (rGO) is prepared by chemical or thermal reduction of GO and has structural defects while maintaining the oxidized groups [89]. For their safe clinical application, their potential biosafety has been assessed by the following studies (Table 2). Chang et al. assessed the in vitro cytotoxicity of GO on A549 cells [70]. The influences of GO on morphology, viability, mortality, and membrane integrity of A549 cells were evaluated by differing particle sizes of 780 ± 410, 430 ± 300, and 160 ± 90 nm, respectively. The cytotoxicity of GO was not significant under 200 µg/mL and did not induce notable cellular uptake. Moreover, the A549 cells maintain their membrane integrity, and normal cell morphology, and did  [71]. Both G and GO exhibited size (small > large > medium), oxidation state, and dosedependent cytotoxicity on cells while having different tendencies on the experimental model used. It was significant that GO induced more DNA damage, intracellular ROS expression, and varied gene expression on HEK 293 T cells than G did, because of its high surface reactivity, which increased their adhering potential for cells and biomolecules. Moreover, small-sized particles on both G and GO showed more severe cytotoxicity because they can be internalized at even lower concentrations.
Other CNM family such as SWCNT, MWCNT, and fullerene is widely applied in the biomedical field using their specific biological characteristics [90,91]. The CNT consists of a layer of graphite rolled up into a cylinder, which is generally divided by SWCNT and MWCNT by their layer arrangement. Their excellent electrical conductivity makes them have a great affinity for bioactive molecules like DNA, protein, cells, and microorganisms [92]. However, the reduced sizes of CNTs induce increment in contact area with the cellular membranes leading to unwilling intercalation and superoxide/ROS formation in the cells. Sayes et al. aim to enhance the cytocompatibility of SWCNT by surface modification using -SO 3 H, -SO 3 Na, and -(COOH) 2 coatings [72]. The live/dead assay and immunofluorescence staining on NHDFs were assessed at a concentration under 2000 µg/mL at 24 h culture. The SWCNT-phenyl-SO 3 H (carbon/functional group ratio = 18) showed excellent cytocompatibility up to 2000 µg/mL, while other types of surface-coated SWCNTs and surfactant stabilized ones exhibited severe cytotoxicity under 20 µg/mL. This suggests that as the degree of sidewall functionalization increases, CNTs become more cytocompatible by preventing direct contact with cells, which can be observed in precipitants in surfactant-stabilized groups. To elucidate the cytotoxic mechanisms of MWCNTs, Patlolla et al. studied the effect of MWCNTs on NHDFs in terms of cytotoxicity, genotoxicity, and apoptosis ( Fig. 3) [73]. During 96 h of culture, cell membrane integrity and viability were continuously decreased in 40, 200, and 400 µg/mL groups. Despite the existence of COOH groups on the surface, the DNA damage and the number of apoptotic cells were dose-dependently e Relative gene expression of wound healing-relative genes. f in vivo toxicity using Caenorhabditis elegans models. The survival rates were evaluated by microscopic observation. The asterisks (* ~ ****) indicate significant difference (*p < 0.05, **p < 0.01, ***p < 0.001, and ****p < 0.0001, respectively.). The data was reproduced from ref [78]. Copyright 2022 MDPI increased indicating their severe genotoxicity. Therefore, it is suggested that the toxicity at high concentration should be carefully considered which is essential for risk control. Another study conducted by Liu et al. focused on the lengthdependent cytotoxicity mechanisms of MWCNTs on RAW 264.7 and MCF-7 cells [74]. The results showed that the 3-14 µm (longer) MWCNTs induced high toxicity, especially to RAW 264.7 cells, while 1.5 µm (shorter) MWCNTs were relatively less cytotoxic. This can be contributed to the enhanced phagocytosis by immune cells as well as proinflammatory responses by MWCNTs, proposing the intercalation rate of CNTs in cells could be one of the important factors for determining cytotoxicity.
A carbon fullerene is described as 60-liked carbon atoms in a highly stable icosahedron [93] and its water-soluble form has been highlighted for biomedical application by its several biological effects such as activities against cells, enzymes, DNA molecules, and free radicals [94]. The cytotoxicity of water-soluble fullerene in HUVECs has been demonstrated by Tamawaki et al. [75]. The 100 μg/mL hydroxyl fullerene (C 60 (OH) 24 ) was treated for 24 h and led to cytotoxic morphological changes such as cytosolic vacuole formation and lowered cell density in a dose-dependent manner by damaging membrane integrity and hampering proliferation. The C 60 OH 24 caused the accumulation of polyubiquitinated proteins that facilitated autophagic cell death. The chronic treatment of C 60 OH 24 (10 µg/mL for 8 days) inhibited cell attachment and growth. As described above, different geometric structured CNMs induce their cytotoxicity and need to be directly compared to each other. Jia et al. compared the three types of CNMs which are fullerene, SWCNT, and MWCNT on alveolar macrophages. The cytotoxicity showed a mass-based sequence order (SWC-NTs > MWCNT > quartz > C 60 ). Compared to MWCNT and fullerene, SWCNT impaired the phagocytosis of macrophages at a relatively low concentration (0.38 μg/cm 2 ). On the other hand, SWCNT and MWCNT of 3.06 µg/cm 2 induced cell necrosis, degeneration, and chromatic condensation that could be a consequence of the apoptotic process [76].
The in vivo toxicity of CNMs generally has been performed on animal models such as mice and rats. Using the animal test, biodistribution, clearance, hematology, and histopathology are widely assessed. Because the localized CNMs in the lung, eyes, brain, liver, kidneys, heart, and spleen can induce complex toxicity and side effects, the in vivo toxicity of CNMs should be sophisticatedly evaluated before clinical administration [95]. Chong et al. evaluated the in vivo biocompatibility of PEG-Cy7-functionalized GQD [77]. The 200 μL of PEG-Cy7-GQD was i.v. and i.p. injected at a concentration of 1.5 mg/mL to 4T1 tumor-bearing BALB/c mice and observed for 48 h. The PEG-Cy7-GQD showed enhanced biocompatibility than GO owing to its small size and high oxygen contents. After multiple times of injections, PEG-Cy7-GQD showed no accumulation in main organs and fast clearance through the kidney, while none of both i.v.-and i.p.-injected groups has died for 54-day post-injection. It was suggested that PEG-Cy7-GQD did not hamper mice owing to its small size, while GO exhibited severe toxicity that led to mice death due to aggregation inside the organs. Salesa et al. assessed the in vitro and in vivo toxicity, cell proliferative activity, and gene-modulating effects of multi-layer G (approximately 2 to 10 layers), also called GNPs, using HaCaT cells and Caenorhabditis elegans model (Fig. 3) [78]. The GNPs exhibited EC 50 of 1.142 µg/mL and 0.760 µg/mL at 12 h and 24 h, respectively, showing time-and dose-dependent cytotoxicity on HaCaT cells. Moreover, GNPs treatment at HaCaT cells increased proliferation and upregulated wound healing-related genes. However, the in vivo administration in Caenorhabditis elegans showed both acute (24 h) and chronic (72 h) toxicity at 25 and 12.5 µg/mL, respectively, suggesting that low doses should be maintained for clinical applications of GNPs. On the other hand, CNTs often contaminated during the growth and fabrication process by disordered carbon, metallic impurities, and polycyclic aromatic hydrocarbons. Koyama et al. focused on the immunological toxicity of CNT impurities to avoid misunderstanding in the interpretation of its in vivo toxicity [79]. The 1 mg of MWCNTs were i.d. injected into the BALB/c mice and immunohistochemical analysis was conducted for up to 4 weeks. The results indicated that MWCNTs with impurities induced immunological toxicity and localized alopecia, whereas pure MWCNTs showed excellent biocompatibility. Therefore, it is suggested that thermal treatment before the usage of MWCNTs in vivo could be a helpful way to improve their biocompatibility. Tang et al. fabricated PEGylated MWCNT (s-MWCNT-PEG) with short-length (5-200 nm) and evaluated its in vivo biosafety [80]. The s-MWCNT-PEG did not induce inflammatory responses, the coagulation system, hemograms, or vital organ functions. After long-term observation, there were no hampering effects on male mouse sperm production or mutagenesis suggesting its biocompatibility.

Biosafety of metal-based nanotheranostics
Metal NPs have gotten tremendous attention thanks to their specific and intrinsic chemical, biological, and physical properties [104,105]. Especially, numerous efforts have been made to explore the theranostic properties of AgNPs in practical applications including anti-bacterial and anti-cancer therapeutics, diagnosis, and treatment of burn wounds, scalds, ulcers, and conjunctivitis [106,107]. A special interest of AgNPs is focused on their antimicrobial-related usages for diminution or deprivation of their anti-pathogenic activity [108]. AgNPs induce microbial membrane damage by a serial process of physical attachment, membrane piercing, and cytoplasm leakage [109,110]. Furthermore, the release of free Ag + ions leads to intracellular ROS generation and the inactivation of essential macromolecules such as proteins, enzymes, and nucleotides [111]. This phenomenon could be applied in mammalian cells; therefore, the cytotoxicity of AgNPs has been assessed by several works (Table 3). Mukherjee et al. compared the in vitro cytotoxicity of AgNPs between different cell lines [96]. The 65-to 69-nm sized spherical AgNPs were treated up to 1000 µg/mL 72 h on HaCaT cells and HeLa cells. HeLa cells were found to be more sensitive than HaCaT cells in all the assays, which can be attributed to their different intrinsic antioxidant levels. The treated AgNPs induced elevated levels of oxidative stress, GSH depletion, and damage to the cell membrane that leads to apoptosis. Moreover, the interaction between AgNPs and the culture medium indirectly poses cytotoxicity. Therefore, it is suggested that the natural defense mechanisms on oxidative stresses of different cells could be a key point to decide the treatment of AgNPs. On the other hand, Grosse et al. assessed the neuronal degeneration and necrosis of RBE4 cells induced by AgNPs [97]. The spherical citratecoated 10-to 100-nm sized AgNPs were treated a 1-25 µg/ mL for 24 h. The results indicate that the cytotoxicity of AgNPs is related to particle size, surface area, dose, and time. The exposure to Ag + ion induced less damage to the cells than AgNPs suggesting the release ions are not a dominant factor for cytotoxicity. Meanwhile, surface stabilization and functionalization are known to be effective methods to enhance the cytocompatibility of AgNPs. Vuković et al. introduced AOT, PVP, PLL, and BSA stabilization on the 7-to 21.1-nm sized spherical AgNPs and treated to hPBMCs at a concentration of 1-25 µg/mL for 3 h (Fig. 4) [98]. The AgNPs altered morphology of hPBMCs inducing apoptosis and cell death via ROS production and mitochondrial damage. Within the protein coatings, positive-charged protein coating induced the highest cytotoxicity in a dose-and timedependent manner because of the favorable internalization.
AuNPs also have been highlighted in the biomedical field owing to their specific characteristics such as high light absorption coefficient, ease of tailored physicochemical modification, strong binding property to amino acids, and tunable optical and electrical properties [112][113][114][115]. Especially, the light absorbance of AuNPs can be tuned from visible to near-infrared (NIR) region by changing size and morphology. Because the NIR window lights can penetrate human tissue, an optical-based application such as photobiomodulation therapy using AuNPs in deep tissues is effective [116,117]. Moreover, the electromagnetic field enhancement at the edge of the geometric AuNPs makes them be used in surface-enhanced Raman spectroscopy [118]. Besides these characteristics, AuNPs are combined with biomolecules and drugs to be applied in therapeutics [119]. Although the biosafety of AuNPs is relatively well studied, several controversial results and their potential toxicity have been demonstrated.
The AuNPs can be fabricated into unique geometric conformations such as nanorods, nanowires, nanocages, nanostars, and nanodendrites, which endow them with versatile properties [120]. Woźniak et al. demonstrated the shapeand size-dependent in vitro cytotoxicity of AuNPs in different conformations [99]. Based on WST-1 assay and live/ dead assay, cytotoxicity of spherical (~ 10 nm), nanoflowers (~ 370 nm), nanorods (~ 41 nm), nanoprisms (~ 160 nm), and nanostars (~ 240 nm) on HeLa cells and HEK293T cells was elucidated. The 1-300 µM AuNPs were treated for 72 h, suggesting nanospheres and nanorods were more toxic than the star, flower, and prism structures. This could be contributed to enhanced cellular endocytosis and aggregation of small-sized nanoparticles, which suggest that particle size is a more dominant factor than particle shape. Chuang et al. used a cell-impedance system which is a dye-free and real-time screening platform by measuring changes in the cell's electrical impedance [100]. To assess the reliability of the assay, 39-to 45-nm sized nanorod AuNPs were treated to AGS cells, A549 cells, NIH3T3 cells, PK-15 cells, Vero cells, and MRC5 cells in the concentration of 1-1000 µg/mL for 72 h. The results show that the nanorod AuNPs induce size-and dose-dependent cytotoxicity by altering signaling and gene expression that led to apoptosis or hampering the cell cycle. The degree of cytotoxicity differed by cell type and the results were comparable to traditional methods suggesting the cell-impedance method could be a reliable way to measure dynamic cell viability.
Yen et al. compared the cytotoxicity and immunological response of AuNPs and AgNPs of different sizes [101]. The treatment of either type of NP at ≥ 10 ppm dramatically decreased cell viability. Especially, small-sized AuNPs upregulated expressions of proinflammatory genes IL-1, IL-6, and TNF-α. It is suggested that negatively charged AuNPs absorb serum proteins and are endocytosed via a complex pathway, which results in higher cytotoxicity and immunological responses compared to AgNPs. Despite the wide usage of AgNPs in wound dressings and other medical applications, their potential in vivo toxicity has not yet been precisely elucidated. In the biological system, AgNPs can lead to free radicals and ROS generation, cell apoptosis, and chromosomal aberration [121]. Therefore, the in vivo biosafety of AuNPs was assessed by Tiwari et al. [102]. The 13 ± 1-to 135 ± 1-nm sized spherical AgNPs were i.v. injected with 4, 10, 20, and 40 mg/ kg to Wistar rats. Based on the biochemical panel assay and evaluation of histochemical parameters, it was concluded that the AgNP in doses (< 10 mg/kg) is safe and has no side effects, but its high dose (> 20 mg/kg) may induce several toxic effects. On the other hand, Chen et al. evaluated the in vivo biocompatibility of AuNPs and analyzed detailed alterations of behaviors  [103] and morphologies of mice [103]. In sizes of 3-100 nm, there were no harmful effects; however, from 8 to 37 nm induced severe sickness in mice. Sizes 8-37 nm AuNPs induced fatigue, loss of appetite, change of fur color, and weight loss in mice, which finally died within 21 days. Moreover, the increase of Kupffer cells in the liver, loss of structural integrity in the lungs, and diffusion of white pulp in the spleen were induced, mainly by antigen-antibody responses. Using the zebrafish model, Ramachandran et al. compared in vivo toxicity of AgNPs and AuNPs [122]. Both particles are up to 30-nm sized spherical forms and 9.7-58.2 µg/mL was administered to zebrafishes. After 96 h, LC50 values of AgNPs and AuNPs were 25 µg/mL and 40.1 mg/mL, respectively. Half of the LC50 was exposed to zebrafishes for 14 days and the following alteration of organs were observed. The AgNP-treated gill and liver tissues showed morphological changes such as cell membrane damage, irregular cell membrane, karyopyknosis, and complete disruption of the gill. In contrast, AuNP-treated fishes showed such changes in only livers; however, AgNPtreated liver cells showed higher levels of ROS generation than AuNP-treated ones. Moreover, AgNP-treated blood cells showed micronuclei formation and abnormal nuclei morphology, while there were no such effects on AuNP-treated ones. Therefore, it is suggested that AuNPs and AgNPs have different toxic mechanisms, hence, should be carefully considered before administration.

Transition metal carbides, nitrides, and carbonitrides (MXene) nanotheranostics
In the field of bioimaging and therapeutics, inorganic 2D nanotheranostics have been intensely developed since the . d ROS production and mitochondrial membrane potential of hPBMCs after 1 (white columns) and 3 h (grey columns) of exposure. The asterisk (*) indicates significant difference (p < 0.05). The data was reproduced from ref [98]. Copyright 2020 MDPI discovery of G nanomaterials owing to their versatile characteristics such as high aspect ratio, unique surface chemistry, unique optic properties, and quantum size effect. Vast kinds of 2D nanomaterials have been developed including transition metal dichalcogenides (TMDs) [128,129], hexagonal boron nitride [130,131], 2D metal oxides/sulfides [132,133], MXenes [134], 2D metal-organic frameworks (MOFs) [135], and 2D perovskites [136]. Because of the large extent of the metal-based 2D nanotheranostics, MXene-based 2D nanotheranostics was focused in this paper. 2D layered MXene, synthesized by selective etching of A elements from three-dimensional (3D) MAX phases to have a metallic M-A bond and a mixed M-X bond, provides unique excitation-dependent fluorescence properties and high drug loading capacity [137,138]. Although the cytotoxicity of MXene is reported to be mild compared to other types of metal-based 2D nanomaterials, their cytotoxic potential and biocompatibility should be discussed for safe clinical application (Table 4). Scheibe et al. fabricated multi-, few-, and single-layered Ti 3 C 2 T x MXene nanosheets, as well as TiC, Ti 2 AlC, and Ti 3 AlC 2 , and compared cytotoxicity, membrane permeability, ROS stress, and mechanical stress derived from different morphology and size (Fig. 5) [123]. The two different cell lines MSU1.1 and HeLa models showed different cytotoxicity results to their tumorigenicity. The results indicated that exposure to higher concentrations of the particles with sizes < 44 μm could be harmful to cell membrane disruption for TiC and MAX phase. Meanwhile, every form of Ti3C2Tx showed cytocompatibility to MSU1.1 cells with only slight cytotoxic behavior in the highest concentration (400 µg/mL). The MXenes exhibited a strong affinity to cell actin filaments to induce cytoskeleton changes but did not impact the cytotoxicity. Overall, cytotoxicity of MXene nanosheets was observed for cells of cancer origin suggesting their cell-dependent working mechanisms. Lee et al. assessed the cytotoxicity of Single-or few-layered Ti 3 AlC 2 MXene nanosheets on MC3T3-E1 cells [55]. Using the CCK-8 assay, there was no significant cytotoxicity under 125 µg/ml MXene treatment; however, time-and dosedependent cytotoxicities were observed at 250 µg/ml after 24 h and 62.5-250 µg/ml after 48 h. This cytotoxicity is suggested to be derived from serious membrane damages at 250 µg/mL, which was assessed from LDH release assay and microscopic observation. Zhang et al. revealed that there was no obvious acute cytotoxicity to HUVECs via a combined untargeted and targeted metabolomics approach [124]. After 100 and 500 μg/mL Ti 3 C 2 MXene nanosheets were treated on cells, SEM observation indicated that MXene nanosheets contacted cell membranes and formed spherical structures while the number of live, apoptotic, and necrotic cells was comparable between control and MXene-treated groups. However, at 500 μg/mL MXene-treated groups, HUVECs showed enhancement of glycolysis, fatty acid biosynthesis, and lipid accumulation suggesting the metabolic shift has been provoked.
Nasrallah et al. estimated the potential in vivo acute toxicity of Ti 3 C 2 T x MXene nanosheets using the zebrafish embryo model [125]. The results indicate that there was no significant cumulative mortality up to 50 μg/mL. However, the cumulative mortality rate was shown to be 21% at 100 μg/mL and it was dose-dependently increased at higher concentrations. After 96 h after fertilization, the LC50 was observed at 257.46 μg/mL, showing that MXene-treated embryos have normal locomotion behavior at 50 μg/mL, nor did Ti3C2Tx nanosheets at this concentration hindered neuronal or muscular activities of the zebrafish embryos. Alhussain et al. evaluated in vivo toxicity of Ti 3 C 2 T x MXene nanosheets in the early stage of chicken embryos [126]. After 5 days of incubation, 30 µg MXene nanosheets per single chicken embryo induced severe toxicity leading to 46% death. Moreover, significant inhibition of angiogenesis in the chorioallantoic membrane was observed by microscopic analysis. Several genes including ATF3, FOXA2, INHBA, SERPINA3, and VEGF-C were down-regulated, which control cell proliferation, survival, cell death, and angiogenesis. These findings suggest that MXene nanosheets potentially exhibit toxicity in the development stage of embryos.
Lin et al. assessed the in vivo biosafety of Nb 2 C MXene nanosheets for their photothermal tumor eradication application [127]. The prepared Nb 2 C MXene showed no significant cytotoxicity at 0-200 µg/mL concentration on 4T1 and U87 cells up to 48 h of treatment. In Kumming mice, measurement of normal hematology parameters showed there was no meaningful changes from the Nb 2 C-treated groups indicating there was no inflammation and infection. The standard blood biochemical indexes indicated that there was no abnormities in blood biochemical indexes and no significant renal and hepatic toxicity. In terms of biodegradation, 20% of Nb content was excreted out by urine and faeces after 48-h post i.v. injection. Moreover, the H&E staining during 28 d suggested that there was no significant acute, chronic pathological toxicity and adverse effects on main organs. The biodegradable properties of Nb 2 C MXene nanosheets are comparable to other kinds of rapidly biodegradable 2D inorganic nanotheranostics such as silicene [139] and germanene nanosheets [140] suggesting its excellent clinical potentials.

Iron oxide-based nanotheranostics
In the last decades, ceramic nanomaterials comprising iron oxide, hydroxyapatite (HAp), zirconia (ZrO 2 ), silica (SiO 2 ), titanium oxide (TiO 2 ), and alumina (Al 2 O 3 )   [127] have been used in theranostic application due to their controllable drug delivery, high stability, inertness, and the feasibility of diversified administration routes such as oral and inhalation [147]. Within the ceramic-based nanomaterials, several types of iron oxides, mostly maghemite (γ-Fe 2 O 3 ) and magnetite (Fe 3 O 4 ), with proper surface chemistry have been intensively used material for in vivo applications such as MRI contrast agents, tissue engineering scaffolds, immunoassays, hyperthermal probe, and drug delivery carrier [148][149][150]. Especially, SPION-based MRI enables the diagnosis of various diseases including oncological pathologies without using radiation or harmful radiotracers [151]. The high ligand affinity of IONPs allows tuning of pharmacokinetics, as well as a combination with fluorochromes and radioisotopes for detection with nuclear imaging such as PET and SPECT [152]. In terms of biosafety, iron is one of the most abundant elements in the body system and plays an essential role in cellular respiration and oxygen transport. Therefore, body-injected iron oxide NPs can be incorporated into natural metabolic pathways of the human body and have high biocompatibility and biological tolerances [153]. However, despite their relative non-toxicity, iron oxide can contribute to cytotoxicity by ROS generation; hence, surface functionalization has been widely applied for their safe application (Table 5). Poller et al. evaluated the in vitro cytotoxicity of different surface-functionalized SPIONs on various cells (Fig. 6) [141]. The Dex and LA have been coated on 44.6-78.9 nm spherical SPIONs and treated on BT-474 cells, T-47D cells, MCF7 cells, MDA-MB-231 cells, and HUVECs at the condition of 0-75 µg/mL for 48 h. The results show that every SPIONs showed dose-dependent cytotoxicity, and the breast cancer cells and endothelial cells exhibited a different response to each SPIONs. Especially, the internalization by cells strongly related to the surface coating (LA-HAS and Dex < LA) induces the most severe cytotoxicity on LAcoated SPIONs. Therefore, it is suggested that the coating materials of SPIONs can be differently selected by their purposes as therapy or diagnosis. The uptake kinetics of SPION are highly related to surface chemistry. Hong et al. assessed the cytotoxicity and genotoxicity of SPIONs coated with various functional groups [142]. The SPIONs were modified with hydroxyl, carboxylic, and amine functional groups, along with surface coating using TEOS, APTMS, and citrate. Within the concentration range of 0-1000 µg/mL, mitochondrial activity, intracellular accumulation of reactive oxygen species, membrane integrity, and DNA stability of L-929 cells were evaluated. Up to 1000 µg/mL, there was no significant cytotoxicity, membrane damage, or intracellular ROS generation in every group; however, DNA damages were observed at high concentrations of APTMS, TEOS-APTMS, and citrate-coated groups. The TEM images cells indicated that larger and negatively charged SPIONs are more actively internalized into cells and lead to damage of intracellular vesicles. Hanot et al. assessed different uptake kinetics of CHO-K1 cells on different surface-coated SPI-ONs [143]. The starch-coated, aminated, and PEGylated 50to 100-nm sized spherical SPIONs were treated to CHO-K1 cells at 0-2000 µg/mL for 72 h. The in vitro assays indicated surface properties of SPIONs have a greater influence on cell uptake than particle sizes. The small-chained PEGylated SPIONs induced lower internalization than longer-chained PEGs, starch-coated, and aminated groups. The internalized SPIONs induced dose-dependent intracellular ROS generation that lead to severe cytotoxicity, suggesting the avoidance of particle uptake can be an effective way to reduce cytotoxicity.
Meanwhile, the in vivo toxic effects of IONPs were demonstrated by Szalay et al. [144]. The spherical IONPs under 50 nm sizes were administered to Wistar rats by intratracheal instillation at 1-5 mg/kg and biocompatibility was assessed by histopathology and bacterial reverse mutation assay for 4 weeks. The results show that there were no pathological changes in the internal organs of mice except for weak pulmonary fibrosis and none of the mutagenic effects in the bacterial system was evoked, suggesting the excellent in vivo biocompatibility of IONPs. In the other study conducted by Prodan et al., IONPs under 21 nm sizes were administered to Brown Norway rats by i.p. injection [145]. After 48 h, the histopathology shows that there was no modification on the liver, kidney, lung, and spleen, indicating the excellent biosafety of IONPs. For IONPs, the inappropriate surface coating can reduce biocompatibility by facilitating cellular uptakes in vivo. Feng et al. fabricated PEI-coated and PEGylated IONPs with sizes of 17.2-35.8 nm and assessed in vitro and in vivo biocompatibility (Fig. 7) [146]. In in vitro assay using RAW 264.7 cells and SKOV-3 cells, the size, dose, and surface coating-dependent toxicity were observed 24 h after treatment. Especially, cellular uptake was most actively induced in PEI-coated IONPs, which lead to severe cytotoxicity by ROS production, autophagy, and apoptosis. On in vivo assays, some of the particles accumulated in the liver and spleen, and finally slowly cleared after 2-week post-injection. Consequently, the dose, particle size, and surface chemistry should be controlled to ensure the safe clinical application of IONPs.

Silicon-based nanotheranostics
Silicon, or Si, constitutes the most abundant component of the Earth's crust. Nanoparticles made from silicon-based materials and their oxides are widely investigated for biomedical applications since it is possible to control the size, porosity, and shape for specific biological applications possessing a relatively inert chemical composition. Siliconbased nanoparticles are promising probes for bioimaging, photodynamic therapy, radiotherapy, and various biomedical applications [158][159][160]. SiNP can be easily synthesized, and be modified in the form of surface modifications, which allow for the conjugation of targeting molecules. Such targeting molecules have been shown to independently affect the targeting, drug loading, circulation, and cellular uptake properties of SiNP. Once the target is recognized, nanoparticles can release their drug payload in a controlled manner by tailoring the interior structure of the particles for desired release profile [161,162]. SiNPs also benefit from favorable biodegradability and the ability to store both hydrophilic and hydrophobic drugs. For these reasons, SiNPs have been extensively used in the preclinical stage and have been used in many diagnostic and therapeutic applications, including photodynamic therapy, imaging, drug delivery, gene transfection, DNA, and microarray detection [163,164]. However, the knowledge of its adverse toxicity and related mechanisms is still limited, and it is important to distinguish how the variations of surface modification and size can influence the outcoming toxicity. This section explores the current research on the toxicity of SiNPs. Summaries of experimental conditions and results are provided in Table 6. Moghaddam, S. P. H., Mohammadpour, R., and Ghandehari, H. fabricated five SiNPs with differences in size, porosity, density, and composition and tested them for in vitro and in vivo toxicity, degradation, biodistribution, and clearance (Fig. 8) [156]. For investigating intracellular degradation of SiNPs, they were added RAW 264.7 macrophages. Cells treated with 20, 40, and 80 μg/mL of Meso-100 (the number indicates the diameter of SiNPs, nm) particles showed 67%, 50%, and 21% of cell viability while other groups showed > 70% of cell viability under the same concentrations. In vivo degradation demonstrated that more porous and smaller particles with higher surface areas have faster degradation rates. The degradation percentage values of Meso 100, Meso 500, and Stöber 100 (Stöber means the particle group made from Stöber method) nanoparticles were ca. 3.8, 2, and 0.3%, respectively. In addition, pH-dependent degradation was observed and nanoparticles in the medium pH value bigger than 4.5 had higher degradation values. Biodistribution results showed all particles accumulated in the liver and spleen more than kidney and lung. These integrated results indicate that the size and porosity of the SiNPs play predominant roles in cytotoxicity and degradation. Elbialy, Nihal S., et al. developed MSNP with PEG coating to endow immune system avoidance containing curcumin as a nutraceutical anticancer agent [157]. For investigating the in vitro anticancer activity, PEG-MSNPs-Cur was treated on HepG2 and HeLa cells. PEG-MSNPs-Cur exhibited higher cellular uptake and significant cytotoxicity against cells compared to free curcumin. Flow cytometric experiment demonstrated that nanoparticle-treated cancer cells showed marked cell cycle arrest at G2/M. In vivo, anticancer efficacy was evaluated, and tumor growth inhibition was the most significant in the PEG-MSNPs-Cur group (~ 80%). The toxicity of PEG-MSNPs-Cur was assessed by histopathological examination of the heart, liver, kidney, and spleen to address the biosafety issue of PEG-MSNPs-Cur. All of the organs showed normal morphological shape with no defects indicating that the application of PEG-MSNPs-Cur as anticancer drugs is nontoxic and safe on normal tissue. Park et al. investigated the cytotoxicity of SiNPs on ocular surface cells such as HCECs [154]. Cytotoxicity induced by 50-, 100-, and 150-nm sized SiNPs was evaluated and there was no significant toxicity in HCECs cultured with all three SiNPs sizes. Cellular uptake and intracellular distribution of SiNPs were evaluated. The nanoparticles were localized mainly in the cytoplasmic matrix without mitochondrial damage or nuclear membrane damage. ROS assay revealed that intracellular and extracellular ROS levels of HCECs were dose-dependent manner and ROS increase by smaller particles was significantly higher than that by bigger particles. However, through the western blot assay, they found that intracellular survival machinery such as the mTOR pathway and cellular autophagy remained activated. -PEI coating induced higher cellular uptake that caused severe cytotoxicity through ROS production, autophagy, and apoptosis -Accumulated in liver and spleen, which showed slow (> 2 weeks) clearance [146] These integrated results indicate that 50-, 100-, and 150nm sized SiNPs are relatively safe at concentrations below 100 μg/mL. The toxicity of SiNPs to the immune system is important in safety considerations, but the related mechanisms are still limited. Immune cells such as phagocytes, dendritic cells, and T-lymphocytes play an important role in host exposure to nanoparticles. They uptake and process NPs, but cell functions change after exposure mediated by activated inflammatory signaling pathways and oxidative stress. Heidegger et al. assessed the biocompatibility of MSNPs in primary murine immune cells [155]. The MSNPs were rapidly taken up into the endosomal compartment by dendritic cells and showed a favorable toxicity profile and did not affect the cell viability in relevant concentrations (100 µg/ mL). Additionally, MSNPs induced only very low immune responses in cells which is determined by the low expression of activation markers of pro-inflammatory cytokines such as IL-6, -12, and -1β.

Biosafety of quantum dots (QD)
QDs are semiconductor nanocrystals that possess unique luminescent properties such as high fluorescence, resistance to photobleaching, broad photoexcitation, and tunable luminescent color with the size ranging from 2 to 10 nm The asterisks (* ~ ***) indicate significant differences (*p < 0.05, **p < 0.01, and ***p < 0.001). The data was reproduced from ref [141]. Copyright 2017 PubMed Central (PMC) [179]. QDs have been considered as promising biomedical-imaging and drug delivery agents. QDs for biomedical application typically have a core/shell structure and the core of QD is generally composed of toxic elements (e.g., Cd, Pb, Hg) from group II to VI and III to V on the periodic table. Therefore, the crucial factor in toxicity of QD is the stability of core/shell structure to prevent core degradation. Additional coatings or functionalities can be applied to the shells to improve stability or introduce versatile biofunctions . This section explores the current research on the toxicity of QD. Summaries of experimental conditions and results are provided in Table 7.
Historically, extensive in vitro cytotoxicity studies have been done for the use of QDs for in vivo applications. Cytotoxicity is one of the main factors that restrict QDs from the application in cell imaging. QDs can induce different degrees of cytotoxicity, such as reducing cell viability, cell morphology, inducing cells to produce autophagy, and changing gene expression [184]. While Cd-based QDs have the most favorable photoluminescent properties for biomedical applications, they have demonstrated to show obvious cytotoxicity. The Cd-based QDs toxicity mechanism triggers the typical response of cells subjected to oxidative stress from both ROS and cadmium ions [185]. Garmanchuk et al. developed alternative approach for extracellular synthesis of CdS QDs using Escherichia coli bacteria and they measured the viability of HeLa cells after exposed to CdS QDs [165]. Trough the MTT and flow cytometry assay, they found that CdS QDs can inhibit cell growth and cause cell apoptosis and the decrease of cell adhesion. Gholami et al. also observed CdS cytotoxicity of MCF-7 cells and AGS cells [166]. CdS QDs showed significant inhibitory effects on survival rate of both cells in dose-dependent manner, and they had more apoptosis effect on MCF-7 cells rather than AGS cells. Lu et al. tested CdSe/ZnS core/shell QDs induced hepatotoxicity and revealed the underlying mechanism [167]. The hepatic L02 cells were exposed to 5-80 nM QDs for 24-48 h and treatment with QDs reduced cell viability in a dose-dependent manner. Flow cytometry was performed to examine inflammasome activation and pyroptosis in hepatocytes triggered by QDs. The result demonstrated the marked increase of caspase-1 in QDs treated group and confirmed  [146]. Copyright 2018 Springer Nature that the cell death induced by QDs treatment was caspase-1 dependent pyroptosis. Mitochondrial ROS generation was proved to mediate QD-induced caspase-1 increase and QD-induced mitochondrial ROS generation was assessed. mtROS production was dose-dependently increased by exposure to QDs indicating that mtROS mobilization is promising target in preventing hepatotoxicity of QDs. Chen et al. reported that CdTe QDs tended to accumulate in HK-2, HepG2, and MDCK cells, and cause severe toxicity [168]. At the Cd Te concentration of 5 mg/mL in HK-2 cells, 2.5 mg/mL in MDCK, and 10 mg/mL in HepG2, significant toxicity was observed. The LD 50 values for HK-2, MDCK,  [157] and HepG2 were 24.18, 16.32, and 48.04 mg/mL, respectively. The results indicate that the cell lines derived from kidney is more sensitive than cell lines from liver. Jingyasu et al. measured the viability of A549 after exposed to CdSe QDs [169]. At the QDs concentration of 0.5 µM, MTT assay showed the cell viability of 65% as compared to control, and it reduced to 56, 45.1, and 38.01% at 1, 10, and 25 µM of QDs, respectively. Further ROS activity and apoptosis test confirmed the cyto-oxidative response depending upon particle size, and DNA fragmentation analysis showed dosedependent DNA damage. More recently, many researchers proved that QDs could be cytotoxic because they are oxidized while entering the cell, and core ions are released into the cell. Another opinion is that the cytotoxicity of QDs depends on biochemical properties of their surface molecules and there is no such cytotoxicity in nanocrystalline particle itself. Thus, understanding and controlling the role of the surface modification on cytotoxicity is an important topic. Surface modification such as PEGylation can improve aqueous dispersion stability of QDs and sterically hinder the adsorption of opsonizing proteins which led to slow recognition and clearance of particles by the RES. Additionally, targeting molecules, such as antibodies, aptamers, DNA, peptides, or high-molecular weight dextran, can be introduced to the surface coating and allow for versatile biofunctions [186].
Han, Xiao, et al., combined the fluorescence of QDs with SiO 2 nanoparticles having molecularly imprinted polymers (MIPs) layer with bovine hemoglobin (BHb) as a model protein template [170]. Briefly, SiO 2 /CdTe nanoparticles were functionalized with MIP nanolayers by the sol-gel process and PEG chains were grafted onto the surface of the imprinted MIP layer via nucleophilic reaction of the surface amine groups with PEG-NHS, followed by template removal. The nanoparticles were then assessed by detecting BHb in bovine serum. PEGylated group showed better recognition performance. The Ksv values were estimated to be 0.495 and 0.604 µM −1 for the SiO 2 /QDs/MIP and SiO 2 /QDs/ MIP-PEG nanoparticles, respectively. The result indicates that the PEGylation could not only improve the aqueous The data was reproduced from ref [156]. Copyright 2020 Elsevier Inc   [173] solubility of the imprinted nanoprobes, but also greatly increase their fluorescence sensing selectivity. In a recent experiment by Yang, Zhiwen, et al., PEGylated CuInS 2 /ZnS QDs were added to a neuron-like PC12 cells and investigated with neurotoxicity ( Fig. 9) [171]. They found that QDs were taken up by PC12 cells but do not cause significant cell death at QD concentration under 100 μg/mL. In addition, PEGylated QDs obviously inhibited the length and number of neurites of PC12 cells, suggesting that PEGylation of QDs can disturb the signal transmission between nerve cells and reduce the repair ability of nerve cells. These results indicate that although PEG modification endows QDs with many advantages, their potential cytotoxic effects should be more evaluated in their biomedical applications. The major studies of QDs for in vivo models were conducted in mouse or rat. Chen, Tingting, et al. also investigated immunotoxicity of PEGylated CuInS 2 /ZnS QDs using BALB/c mice [172]. BALB/c mice were injected with 1, 5, and 25 mg/kg of QDs via the tail vein. All of the mice survived until day 28, and there were no significant changes ZnS-PEG QDs with a concentration of more than 100 μg/mL in the PC12 cells after 24 and 48 h of incubation (n = 6). d Representative brightfield images of the PC12 cells treated with NGF and 100 μg/mL CuInS 2 /ZnS-PEG QDs + NGF. e Effect of CuInS 2 /ZnS-PEG QDs on the number of neurites of PC12 cells with or without NGF. The data was reproduced from ref [171]. Copyright 2020 Elsevier Inc in body weight or daily behavior. To investigate the biodistribution of mice, immune organs such as spleen and thymus were taken out at different times after mice were intravenously injected with QDs. Fluorescence image showed strong fluorescence signal of QDs from spleen and thymus, and the signals were also observed even after 28 days of injection. H&E staining confirmed that there were no apparent histopathological abnormalities suggesting QDs themselves did not cause any inflammation or histopathological injury to the immune organs. Du, Yan, et al. evaluated in vivo toxicity of PEGylated CdTe QDs and observed the distribution of QDs in Kunming mice [173]. QDs of the doses of 0.4, 2, and 10 mg/kg were intravenously injected into Kunming mice. The body weight of the highest dose group was obviously lower than that in the control group, and inactive state was also observed after 4 days of injection. Organs were dissected to investigate the inflammation and damage of the tissue. H&E-stained slides showed no obvious pathological changes and lesions in the heart, lung, spleen, testis, and ovary treated with high dose of both PEGylated and non-PEGylated group. However, there were obvious pathological changes in the liver and kidney after treatment with high dose of non-PEGylated group. Thus, demonstrating the potential of surface modification with biosafety cues.

Biosafety of molybdenum-based nanotheranostics
Molybdenum (Mo) is essential for nearly all organisms and forms the catalytic center of various enzymes such as xanthine oxidoreductases, nitrate reductases, nitrogenase, and sulfite oxidase. In enzymes, Mo can exist in three oxidation states (+ 4, + 5, and + 6) which make it possible to participate in reduction-oxidation reactions of catalysis. Mo is relatively inert and non-immunogenic. And it has unique mechanical, electrical, chemical, and optical properties in nanoscale. Therefore, molybdenum-based nanoparticles are promising probe for various biomedical applications such as drug delivery, diagnostics, bio-imaging. Mo NPs can generate ROS under ultrasound, and it enables noninvasive treatment using Mo NPs with high penetration depth exceeding 10 cm [177,178]. Although molybdenum has relatively high biocompatibility, it is still necessary to investigate the biocompatibility characteristics according to their size, shape, and crystallinity. This section explores the current research on the toxicity of MoNPs. Summaries of experimental conditions and results are provided in Table 8.
Potential biological response of Mo NPs toward human cells was explored by Akhtar et al. (Fig. 10) [174]. For investigating cytotoxicity of Mo NPs, different concentrations of nanoparticles were treated on human breast (MCF-7) and fibrosarcoma (HT-1080) cells. In all groups ranging from 25 to 800 μg/ml, no significant cytotoxicity was found, and cell viability was similar to control group. To test oxidative stress of nanoparticles in cells, DCF-fluorescence and LDH release were observed. There was no distinct change in DCF-fluorescence and inducement of LDH release in either cell types indicating there is no potential effect of oxidative stress. The ability of molybdenum nanoparticle to induce glutathione in cells was tested. Mo NPs significantly induced glutathione content in cells at concentrations of above 50 μg/  [176]. The cytotoxicity of MoS 2 nanoparticles to the human lung cancer epithelial cells (A549) was tested. In concentration range from 5 to 10 μg/ml, MoS 2 NPs did not produce any significant cytotoxic effect on the A549 cell. However, concentration-dependent toxic profile was observed above 20 μg/ ml, and 60% reduction in cell viability was observed at the concentration of 20 μg/ml. In addition, effect of MoS 2 NPs on the production of ROS was estimated. They found that the ROS generation of MoS 2 NPs was dose-dependent manner. Further flow cytometry confirmed good cellular uptake of MoS 2 by A549 cells.

Challenges
Biosafety of the inorganic nanomaterials is one of the most important factors for their vigorous application as theranostic probes. This review focuses on the categorization of inorganic nanotheranostics by carbon, metal, 2D, ceramic, and semiconductor-based nanomaterials and aims to summarize in vitro and in vivo biosafety studies. Though the toxicity of the inorganic nanomaterials has been elucidated, there are some issues to be resolved.
(1) Most of the toxicity studies have explored sphericalshaped NPs without paying attention to other shaped NPs such as nanorods, nanocages, and nanorings. (2) Lack of human trials limits the clinical application. (3) Different experimental conditions and toxicity evaluation standards lead to controversial results. To address the issues, more clinical trials should be conducted to clarify the long-term toxicity and potential side effects. Real-time monitoring of therapeutic responses of nanotheranostics can provide each patient with immediate feedback on the progress of treatment, allowing the properties of nanotheranostics to be modified as needed. Moreover, we are now moving toward personalizing treatment according to the patient's requirements, which also requires generating and processing huge amount of information using artificial intelligence (AI). For personalization, the concept of combination nanomedicine is also needed. It involves the synergistic dynamics of one or more therapeutics which promise to overcome challenges such as development and optimization of nanomedicine. Moreover, novel systematic toxicity assessment models are required to elucidate the intrinsic mechanisms of toxicity of inorganic nanomaterials using various types of cells and animals. We expect that the researchers with interdisciplinary backgrounds will advance the field of inorganic nanomaterial-based theranostics by reflecting on the problems and the concerning suggestions.

Conclusions and perspectives
In summary, inorganic nanomaterials have various elemental and chemical formulations of which the cytotoxicity can be tailored for different theranostic purposes. CNMbased nanotheranostics show encouraging cell-imaging applications. They enter cells without any further functionalization, and the fluorescence property of CNM-based particles can be used for fluorescence-based cell imaging applications. Obviously, oxidation and functionalization of G materials induce lower cytotoxicity than pristine G. For CNTs, sidewall functionalization and length-dependence are critical factors for biocompatibility decision. While fullerenes, GNPs, and GQDs have not been extensively studied for theranostic applications, their unique optical properties and high in vivo compatibility promise great potentials. Ag and Au nanomaterials have historically been among the oldest used biomaterials. They have been widely used in diagnosis and detection application due to their specific properties such as large specific surface area, strong localized surface plasmon resonance effect, high extinction coefficient. Also, Ag nanomaterials have been highlighted for their anti-bacterial and anti-cancer effects. Cells tend to exhibit cytotoxicity according to their natural antioxidant levels on Ag nanomaterials treatment. Blood toxicity should be considered because ROS production and mitochondrial damage can be derived from Ag nanomaterials. Au nanomaterials can be prepared in various formulations such as nanoflower, nanorod, nanoprism, and nanostar, and its cytotoxicity is mainly determined by surface chemistry and aggregation properties. However, since high doses of Ag and Au nanomaterials can lead to main organ damage or mortality in vivo environment, careful dose determination is required. The biosafety of versatile types of inorganic 2D nanotheranostics is also discussed. Within last 10 years, TMDs, hexagonal boron nitride, 2D metal oxides/sulfides, MXenes, 2D metal-organic frameworks (MOFs), and 2D perovskites have been synthesized and their biomedical application extensively studied. Within these, the cytotoxicity of MXene-based nanomaterials was discussed suggesting their relatively high cytocompatibility, hemocompatibility, and excellent biodegradation properties. MXenes have many biomedical applications such as biosensing, bioimaging, and drug delivery. Especially, it can be applied to photodynamic therapy owing to the photothermal effect of MXenes. The ceramic-based inorganic nanomaterials such as iron and silicon-based nanomaterials feature high biocompatibility due to their bio-inertness, hence, can be used for long-term in vivo imaging applications. Semiconductor nanotheranostics such as Cd, Se, Zn, and Te compounds are also utilized in bioimaging because of their unique luminescent properties such as high fluorescence, resistance to photobleaching, broad photoexcitation, and tunable luminescent colors. Although potential biosafety should be further elucidated, their excellent optical characteristics promise great potential as future theranostic probes. Lastly, molybdenum nanotheranostics are relatively biocompatible, and it has unique mechanical, electrical, chemical, and optical properties which make it possible to use as various biomedical applications. Especially when used with ultrasound, Mo NPs can generate ROS, and it enables noninvasive treatment with high penetration depth. Although molybdenum is a material with high biocompatibility, it seems that more advanced research on cytogenetics is needed because toxicity varies depending on particle size or cell type. In this work, we have reviewed the recent advances in inorganic nanomaterials for theranostic application from the aspect of biosafety. Generally, cells can survive to low concentrations (< 10 μg/mL) of nanomaterials for short-term exposure. At the high doses, several groups have dose, time, and surface modification dependent manner of cytotoxic effect. Although many of the nanoparticles' functions are due to their core structure, surface coating defines much of their biological activity. Therefore, the addition of a specific type of surface coating is required for many nanoparticles to be useful for biomedical applications. While nanomaterial-induced cytotoxicity has been reported by several groups, it is important to know that in vitro results may differ from clinical results and are not necessarily relevant. Most of the currently available nanotheranostics have only been investigated in vitro without providing clear evidence for their potential as in vivo imaging and therapeutic agents. The therapeutic effectiveness of many of the theranostic nanomedicines still needs to be accurately verified. If these hurdles can be overcome, the clinical application of nanomedicine will facilitate the refinement of clinical decisions and treatment protocols by understanding differences among patients and realize the goals of "personalized medicine." Through this review, we expect the contribution to the development of nano-system-based biomedical applications and further studies to better understand the toxicity mechanisms as well as the correlations between nanotheranostics and their impact on human health. While much effort is still needed, developments in the field of nanotheranostics are fast and nanotechnology for personalized medicine will become a reality in the not-too-distant future.
Author contribution All authors contributed to the study conception and design. Dong-Wook Han and Ki Su Kim developed the idea and structure of the paper. The first draft of the manuscript was written by Moon Sung Kang

Competing interests
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.