The toxicity of graphene quantum dots

Abstract

Recently, there has been a rapidly expanding interest in a new nano material, graphene quantum dots, owing to its profound potential in various advanced applications. Despite its exciting application outlook, the toxicology of the material has to be well addressed before its practical use in the highly prospective areas – especially for bio-applications such as bio-sensing, bio-imaging and nanomedicine (e.g. drug delivery). This review provides a comprehensive account of the current research status regarding the toxicity of graphene quantum dots (GQDs), including raw GQDs, chemically doped GQDs and chemically functionalized GQDs. It summarises the existing tests on both in vivo and in vitro toxicity. Important topics including the uptake mechanism by cells and parameters governing the toxicity of GQDs (such as concentration, methods of synthesis, particle size, surface chemistry and chemical doping) are discussed. It also covers demonstrations on toxicity regulation of GQDs via chemical modification, as a toxicity mechanism via generation of reactive oxygen species (ROS) by some GQDs is also evident. Based on the evaluation of the current research status, possible future perspectives are also suggested.

---

Shujun Wang

Shujun Wang is currently a third year PhD student in Queensland Miro- and Nanotechnology Centre (QMNC), School of Engineering, Griffith University. He obtained his bachelor degree in Material Science and Engineering from Beijing Institute of Technology and his Master of Philosophy (Mphil) in Mechanical Engineering from Hong Kong University of Science and Technology where he developed expertise in the preparation of graphene and the application of graphene in transparent conductors. His current research at QMNC of Griffith University involves the synthesis, fundamental studies and applications of graphene quantum dots in sensing and light harvesting.

Ivan S. Cole

Dr Ivan S. Cole is a Chief Research Scientist for CSIRO Materials Science and Engineering Division (CMSE). He joined CSIRO in 1991 as a Research Scientist. Dr Cole was appointed CMSE Deputy Chief, Science in 2007. He fulfilled this role until 2010 and is now a Chief Research Scientist at CMSE. He combines broad experience in the areas of research management and planning with a detailed knowledge of materials science and mathematical modelling. His expertise spans the manufacturing and infrastructure and aerospace sectors.

Qin Li

Professor Qin Li received her BEng and MEng from Zhejiang University, and a PhD in chemical engineering from the University of Queensland, Australia. With a Marie Curie Fellowship, she then worked at the Max Planck Institute for Polymer Research (2006–2009) in the field of advanced nanomaterials. Her faculty career started in 2004 at Curtin University, Western Australia, before moving to Griffith University. She is a leading scientist in the field of fluorescent nanocarbons with a broad interest in functional nanomaterials and nanotechnology for environment and energy.

---

1. Introduction

Graphene quantum dots (GQDs) has rapidly emerged as an important member of the graphene family. As a derivative of graphene, it does not just inherit the excellent properties of graphene but also possesses unique electronic,^1–8 magnetic^9–16 and optical^17–20 properties. Thanks to these properties, GQDs have been regarded as a versatile material which is highly promising in a range of applications, including sensing,^21–37 energy conversion,^38–43 bio-imaging,^31,39,40,44 nanomedicine,^39,45,46 single electron transistors,⁴⁷ spintronics,^9,10 memory^48,49 and so on.

Since its first successful laboratory synthesis, research has focussed on novel synthesis methods, explanation of properties, modification, and demonstrations of potential in different applications. So far GQDs have been synthesized from either top-down or bottom-up routes. Top-down approaches involve cutting certain graphitic precursors, including chemically reduced graphene oxide,^42,50,51 graphene oxide,⁵² graphite,^53,54 carbon fibre,⁵⁵ carbon nanotubes,^56–58 fullerenes, and carbon black,⁵⁹ into small graphene fragments with a lateral size in the nano scale (<100 nm). Bottom-up methods start with organic molecules (i.e. monomers) to assemble dots.^22,25,38,60 Although GQDs produced by these methods vary in size, shape and chemical compositions, the majority of them possess photoluminescence (PL) which puts them into the list of highly prospective candidates for biology-related applications (e.g. bio-sensing, bio-imaging and nanomedicine). Although the exact mechanism of the origin of the PL from GQDs is still a debatable subject, our recent work suggests that it is highly possible that the emission is a combination of individual emission components governed by the carbon core (i.e. sp² carbon), functional groups and defects.⁵⁹ Moreover, due to the rich surface moieties (i.e. functional groups and defects), it is convenient to functionalize raw GQDs by attaching certain foreign moieties covalently^19,30,61 or noncovalently.^46,62 Hence GQDs is an excellent platform which could be tuned for desired applications by chemical modification methods. Apart from raw GQDs and chemically functionalized GQDs, GQDs doped with, foreign atoms, including nitrogen-doped GQDs (nGQDs) and boron-doped GQDs (bGQDs), have also been developed and explored.^63–67

On top of all the excitement about GQDs, the toxicity of GQDs is an essential and critical parameter which ought to be addressed. Unless it is well understood, the practical adoption of the material in real applications, especially for bio-applications, would be difficult. Luckily, there has been increasing effort spent on the toxicological study of GQDs in the past few years. In this review, we will go through the studies to date on the toxicology of GQDs. In particular, both existing in vivo and in vitro tests of GQDs are summarized. The proposed uptake mechanism of GQDs by cells is presented. The parameters which correlate with the toxicity of GQDs are summarized. Certain work done to moderate the possible ROS toxicity mechanism of GQDs via chemical modification is also included. Based on the summary and evaluation of the current research progress, we also suggest some possible future research directions at the end of this review.

2. Existing toxicity tests

Toxicity for living organisms is usually measured and assessed via in vitro and in vivo experiments in a laboratory environment (Fig. 1). The in vitro tests are carried out with living cells from organisms as the test subjects. The in vitro toxicity is also known as ‘cytotoxicity’. In contrast, in vivo tests are conducted with a whole living being as the test subject. Up to now, the toxicity of GQDs has been studied under both in vitro and in vivo conditions. Not just raw GQDs but also doped GQDs (e.g. nGQDs and bGQDs) were included in the toxicity tests. Table 1 summarizes the existing literature that contains toxicity tests as a part or a whole of their studies. So far in vitro tests have been carried out with various human or animal cells. The most studied cells are Hela cells (cervical cancer cells) and A549 cells (lung carcinoma cells); other cells such as MCF-7 (breast cancer cells), red blood cells, and stem cells have also been explored. As for in vivo tests, mice, zebra fish, Caenorhabditis elegans, and green gram sprouts have been adopted in the investigations. In general, there are more in vitro tests available than in vivo tests in the literature.

Fig. 1 Schematic explanation of in vitro and in vivo tests.

Table 1 Summary of testing subjects in toxicity tests of GQDs

Test category	Test subjects	Ref
In vitro	Hela cells (human cervical cancer cells)	68–74
	A549 (human lung carcinoma cells)	71 and 75–77
	MCF-7 (breast cancer cells)	70 and 78
	Red blood cells	79
	MGC-803 (gastric cancer cells)	68
	MHS	80
	Stem cells	81–84
	MCF-10A (human mammary epithelial cells)	70
	THP-1 macrophages	85
	MDA-MB231, MDCK	76
	HEK293	71
	MG-63 (human osteosarcoma), MC3T3	86
	RSC96 (rat Schwann cells)	87
In vivo	Mice	73, 76, 80 and 84
	Zebra fish	70, 74 and 88
	Caenorhabditis elegans	89
	Green gram sprouts	90

---

2.1 In vitro toxicity of GQDs

The cytotoxicity (in vitro toxicity) is typically characterized through cell viability in an in vitro test. In order to obtain the cell viability data, popular testing assays such as MTT, LDH, or ATP are employed. The test material is introduced into the culture environment of the cells and the numbers of living cells at the beginning of the test and after exposure for a certain period of time (known as the incubation time, e.g. 24 h) are measured. The ratio between the two numbers is defined as the cell viability. Apart from the cell viability, researchers also monitor other parameters such as damage to the cell membrane, disruption of morphology, release of certain chemicals due to damage to the cell (e.g. lactate dehydrogenase (LDH), adenosine triphosphate (ATP), lipid extracts…). Table 2 is a summary of the cell viability tests of in vitro tests in the literature. Human cancer cells (e.g. Hela, A549) are the most popular choice in those existing cytotoxicity tests. Fig. 2 shows typical results of the cytotoxicity tests, which indicate that cell viability and common chemical releases from cell (e.g. LDH) are concentration dependant – the viability and chemical releases from the cell gradually declined and rose, respectively, with the increase in concentration of GQDs. Wu et al. did an in vitro toxicity comparison on graphene oxide (GO) and GQDs with the MGC-803 (human gastric cancer) and MCF-7 cells (human breast cancer).⁷⁸ It is obvious from their study that GO has much higher cytotoxicity than GQDs. Nurunnabi et al. compared the cytotoxicity of GQDs among three cancer cell lines – KB (epidermal cancer cells), MDA-MB231, and A549.⁷⁶ In addition to cell viability, they also monitored the intracellular lactate dehydrogenase (LDH) released from the cells as a result of cell membrane damage. They noted that under the same testing conditions (e.g. incubation time, GQDs concentration), the three cancer cells responded to GQDs differently. For instance, MCF-7 cells and A549 are more sensitive to GQDs than the human epidermal cancer cells which had the lowest viability and highest level of LDH release. Apart from cancer cells, the cytotoxicity of GQDs to other human cells has also been evaluated by recent studies. Zhang et al. synthesized GQDs with yellow emission, with which it is possible to image the stem cells that are challenging to label with existing technology.⁸¹ In their toxicity tests, viability above 80% for the chosen stem cells could be obtained after 3 days of culturing with a GQDs concentration of 100 μg mL⁻¹. The study indicates that their GQD is an excellent low-cytotoxicity and biocompatible agent for labelling stem cells. An additional test by Nurunnabi et al. on MDCK (kidney epithelial cells) presented 95% of viability over 48 h while the concentration of GQDs is as high as 500 μg mL⁻¹. In addition to common GQDs, the cytotoxicity of chemically doped GQDs is also among those tested.^68,72,79 Wang et al. studied the cytotoxicity of graphene oxide (GO) and nitrogen-doped GQDs for red blood cells (RBCs).⁷⁹ They evaluated the hemolytic activity and corresponding morphological changes as well as the released ATP of RBCs after separate exposure to GO and nGQDs. This shows that GO causes apparent hemolysis with the release of ATP, but nGQDs do not induce similar damage to the RBCs, which indicates that nGQDs have much lower cytotoxicity than GO. Hai et al. synthesized boron-doped GQDs and tested their biocompatibility with Hela cells.⁶⁸ A 12 h viability of 87% could be reached with an impressively high concentration of 4 mg mL⁻¹ of the boron-doped GQDs. It seems that doped GQDs might even have lower cytotoxicity and better biocompatibility than undoped ones.

Table 2 Summary of in vitro tests^a

Cells for tests	GQDs	Assays	Toxicity	Incubation time	Ref
a Note: nGQD = nitrogen-doped GQDs, bGQD = boron-doped GQDs, via. = viability.
MCF-7, Hela cells, MCF-10A	GQDs	MTT	95% via. at 2 mg mL^−1	24 h	70
Hela, A549 cells	GQDs	MTT, LDH	Via. > 95% at 160 μg mL^−1 (Hela); Via. > 85% at 640 μg mL^−1 (A549)	24 h	73
Hela cells	GQDs	CKK-8	90% via. at 100 μg mL^−1	24 h	74
A549 cells	GQDs	MTT	Via. > 80% at 200 μg mL^−1	24 h	75
KB, MDA-MB231, A549, MDCK cells	GQDs	MTT, LDH	Via. > 95% at 500 μg mL^−1 (MDCK cell)	21 d/24 h	76
A549 cells	GQDs	MTT	80% via. at 100 μg mL^−1	24 h	77
MGC-803, MCF-7 cells	GQDs	MTT	GQDs < GO	3 d	78
Stem cells	GQDs	MTT	Via. > 70% at 200 μg mL^−1	24 h	81
Stem cells	GQDs	MTT	61% via. at 100 μg mL^−1	24 h	83
THP-1 macrophages	GQDs	MTT	82.5% via. at 200 μg mL^−1	24 h	85
MG-63 and MC3T3 cells	GQDs	MTT	Via. > 80% at 400 μg mL^−1	24 h	86
RSC96	GQDs	MTT	70% via. at 200 μg mL^−1	24 h	87
Hela cells	bGQD	MTT	87% via. at 4.0 mg mL^−1	12 h	68
Reb blood cells (RBC)	nGQD	Hemolysis, ATP	nGQD < GO		79

---

Fig. 2 Examples of results of in vitro tests: (a), cell viability for MG-63 cells (replotted with data from ref. 86); (b), cell viability of MCF-7 cells and MGC-803 cells (replotted with data from ref. 78); (c) and (d), cell viability and LDH release of A549 cancer cells (replotted with data from ref. 76).

In general, the majority of the in vitro toxicity tests with different cells and different GQDs suggest that GQDs are indeed a material with low cytotoxicity. However, there is also evidence indicating that the reactive oxygen species (ROS) mechanism, which has been proposed as the toxicity mechanism for other nano materials,^91,92 is not completely absent from GQDs. Markovic et al. found that when their GQDs synthesized from electrochemical exfoliation of graphite rods were subject to blue irradiation, singlet oxygen (i.e. a type of ROS) was generated, leading to the death of U251 human glioma cells.⁸² In their cell viability tests, when the U251 cells are treated with either GQD or blue light alone, no significant cytotoxicity was observed. However, over 40% of the U251 died at a GQD concentration-200 μg mL⁻¹ under blue light irradiation for 24 h, due to the increase in oxidative stress caused by the as-generated singlet oxygen. It was also noted that the cell death took two pathways – apoptosis and autophagy. Another recent study on macrophages also demonstrated that apoptosis and autophagy could be induced by GQDs, owing to the generation of ROS.⁸⁵ This work also showed that ROS exhibited concentration-dependence. In addition, it is also noted that GQDs possess an intrinsic peroxidase-like catalytic activity,⁸⁰ which could even be used to enhance the antibacterial performance of H₂O₂ to treat wound infections.⁹³ Nevertheless, Wu et al. compared the intrinsic intracellular ROS generated by GQDs and GO and revealed that the ROS level induced by their GQDs is lower than that of GO,⁸⁰ but this might not be sufficient to avoid cytotoxicity concerns when specific types of GQDs or cells are under consideration, especially when they are applied in situations with extremely low toxicity tolerance.

2.2 In vivo toxicity of GQDs

The in vivo toxicity tests mainly deal with matters such as the bio-distribution of the testing substance among the organs of a whole living being, the possibility of elimination through the excretory system and any possible damage to organs, etc. In a typical test, GQDs of a certain dosage are injected into the living organism (e.g. a mouse) followed by monitoring the living organism over a period of time (e.g. 24 h). During the period, the bio-distribution, organic accumulation and excretion of GQDs are evaluated. As far as we know, the earliest report containing in vivo work was done by Wu et al. They demonstrated the potential for applying GQDs as synthesized from a bottom-up method for in vivo imaging of mice; however, they did not go further to study the in vivo toxicity of their GQDs.⁸⁰ The first systematic in vivo toxicity study was done a bit later by Nurunnabi et al.⁷⁶ They firstly studied the in vivo imaging, bio-distribution, and ex vivo organic imaging on cancer-bearing mice with GQDs synthesized by exfoliation of carbon fibre. The in vivo imaging shows that 12 h after injection, GQDs could be detected in the tumour site located on the skin, but no fluorescence signal could be picked up from deep organs, which suggests that GQDs could be used for superficial tissue images (e.g. skin cancer detection). 24 h after the injection no GQDs could be detected either at the tumour site or at any part of the mouse body (Fig. 3a). The ex vivo imaging of extracted organs (i.e. kidney, liver, heart, etc.) showed that GQDs were spread throughout the whole body by the circulation system within the first 12 h. During this period, accumulation in different organs evolved, for instance, at the early stage (2 h), GQDs majorly accumulate in the liver and heart. The accumulation in the liver gradually reduced and a significant increase in accumulation at the kidney was observed after 12 h. In general, the fluorescence signal of GQDs at 24 h is weakened in contrast to that at 12 h. The results suggest that GQDs may be conveniently removed by the excretory system. They continued to demonstrate the complete blood count, serum biochemistry analysis and histological analysis of the tissues with a rat model. Within 22 days, no notable difference between the GQDs-injected group and the control group is observed, which leads to the conclusion of the low in vivo toxicity of the GQDs. Recently, in vivo studies were also carried out with zebra fish models.^70,74,88 Wang et al. investigated the influence of GQDs on the development of zebra fish embryos.⁷⁰ They observed that their GQDs majorly accumulate in the intestines and heart of the zebra fish (Fig. 3b). More importantly, it was also learnt that the impact of GQDs on the zebra fish embryos during their development was determined by the concentration. They monitored the embryo growth by varying the concentrations of GQDs (i.e. 50, 100, and 200 μg) and found that a concentration higher than 50 μg mL⁻¹ imposed negative impacts on the embryos of zebra fish, symptoms including a decrease in hatch and heart rate, an increase in mortality, embryo disfigurement, etc. However, another study on zebra fish embryos showed that it is not easy for GQDs to enter the circulation system of zebra fish and the removal of GQDs by the excretory system could be completed in 7 days (Fig. 3c).⁸⁰ No obvious difference between the test group and the control group was found at a concentration lower than 2 mg mL⁻¹, suggesting the excellent biocompatibility of GQDs. They further demonstrated the possibility of applying the GQDs (after functionalization) for in vivo imaging of apoptotic cells to study the apoptosis process. The contradictory results of the two studies on the in vivo toxicity for zebra fish might lie in the different GQDs adopted in their research. The former adopted GQDs synthesized from cutting graphene oxide and the latter used GQDs synthesized from hydrothermal treatment of leaf extracts, which indicates that GQDs synthesized from different approaches have different toxicity features. In addition to animals, there was also work done on plants. Li et al. compared the in vivo toxicity of GQDs with that of three other carbon materials, carbon quantum dots (CQDs), graphene oxides (GO) and single-walled carbon nanotubes (SWCNTs), using growth experiments of green gram sprouts.⁹⁰ The study showed that GQDs are the least toxic material among the four.

Fig. 3 Example of in vivo toxicity tests: (a), mice (reproduced from ref. 76 with permission from American Chemical Society); (b) and (c), zebra fish (reproduced from ref. 70 and 88 with permissions from Elsevier (b) and Royal Society of Chemistry (c) respectively); (d), Caenorhabditis elegans (reproduced from ref. 89 with permission from Royal Society of Chemistry).

In addition to raw GQDs, similar to the in vitro test, an in vivo test was also carried out with nitrogen-doped GQDs.⁸⁹ In this work, the transgenerational toxicity of nitrogen-doped GQDs was examined with both wild and gene-modified Caenorhabditis elegans. They noted that within the tested concentration range (0.1 μg to 100 μg mL⁻¹) nGQDs did not cause death or lifespan decrease, or alter the function of primary and secondary organs. No noticeable damage on the expression patterns of encoded genes was found in the gene-modified Caenorhabditis elegans. The accumulation of nGQD was observed in the intestines of the tested subjects but not in the embryos or progeny. This study not only exhibited the low in vivo toxicity of nGQDs to the elegans but also suggests that the risk of genotoxicity and transgenerational toxicity induced by nGQDs should be very low.

Through the above in vitro and in vivo toxicological studies, in general, GQDs are proved to be a material possessing low toxicity and high biocompatibility. However, GQDs from different synthesis methods have different toxicity profiles and the toxicity to different cells and species is also diverse. In the following we will highlight some important aspects that we have summarized from the existing toxicity researches.

3. Uptake mechanism of GQDs by cells

How GQDs enter cells is directly related to their toxicity, as the process could damage the cell membrane, extract important chemicals and cause various biological responses.⁹⁴ So far, the uptake of GQDs by cells has been commonly attributed to endocytosis^71,78,80,95 (Fig. 4). Wu et al. monitored the internalization of GQDs with human gastric cancer (MGC-803) and breast cancer (MCF-7) cells.⁷⁸ They found that the uptake of GQDs by the selected cells may mainly follow a caveolae-mediated endocytosis and partially involve energy-dependent endocytosis. In addition to cell viability, the intracellular reactive oxygen species (ROS) level, mitochondrial membrane potential and cell cycles were also evaluated, which allowed them to conclude that GQDs have lower cytotoxicity than GO. With human neural stem cells, Shang et al. demonstrated that the endocytosis process for internalization of GQDs is also concentration- and time-dependent.⁹⁵ Moreover, it has also been shown that selective endocytosis could be induced by attaching certain moieties to GQDs, which is very important for application in nanomedicine (e.g. drug delivery).⁷¹ In addition, the majority of the in vivo tests indicate that once inside the cells, GQDs are only present in the cytoplasm rather than in the nucleus and there is as yet no evidence that GQDs could cause destruction to the nucleus.

Fig. 4 Schematic representation of the endocytosis process via which GQDs enter cells.

4. Parameters governing the toxicity of GQDs

4.1 Concentration vs. toxicity

The toxicity of GQDs is concentration dependent. All existing in vitro studies present declining trends in cell viability with an increase in concentration of GQDs in the culture media, as shown in Fig. 2. It was also noticed in the in vivo test of zebra fish embryos⁶⁷ that the in vivo toxicity varied with the concentration. A high concentration (i.e. >50 μg mL⁻¹) leads to an increase in developmental toxicity, which leads to adverse effects in the growth of zebra fish. This is in fact a direct revelation of the dosage dependency of toxicity, which is applicable to most chemicals or materials. Nevertheless, the concentration tolerance of cells to different GQDs is dissimilar. Some GQDs can allow a high cell viability (e.g. above 90%) even at very high concentrations (e.g. 0.5 mg mL⁻¹) for a typical incubation time of 24 h. It seems that most of the studies have a typical ‘cut-off’ concentration ∼200 μg mL⁻¹ only below which high viability exceeding 80% could be obtained.

4.2 Synthesis methods vs. toxicity of GQDs

GQDs in reports incorporating toxicity tests were synthesized from both top-down (e.g. microwave-assisted cutting of GO,⁶⁸ electrochemical unzipping carbon nano tubes,⁶⁹ oxidation of graphite…^73,77) and bottom-up approaches (e.g. pyrolysis of L-glutamic acid⁸⁰ and citric acid⁷⁵ and hydrothermal treatment of pyrene⁸³), as shown in Table 3. Different synthesis methods produced GQDs with dissimilar toxicity traits. Due to the fact that insufficient work has been done on toxicity tests of GQDs from bottom-up methods rather than on those from top-down methods, it is difficult to judge which strategy is superior to the other. Nevertheless, Fig. 5 displays the cell viability values of A549 cells from different reports and picked under the same testing conditions (i.e. a GQDs concentration of 100 μg mL⁻¹ and an incubation time of 24 h). GQDs from pyrolysis of citric acid apparently have a higher viability value than the average of the GQDs from top-down methods, indicating the lowest cytotoxicity and best biocompatibility among those listed. The comparison might imply that GQDs from bottom-up methods might be more biocompatible than those from top-down methods, probably owing to the absence of toxic reactants (e.g. concentrated acid, highly oxidative chemicals) in bottom-up methods in contrast to their popularity in top-down methods, and/or GQDs from top-down ‘cutting’ or ‘unzipping’ methods may have sharper edges, or a more hostile surface chemistry than those from bottom-up approaches, making them more destructive to cells. Such an implication does need further verification.

Table 3 Summary of synthesis routes for GQDs in different toxicity tests^a

Synthesis category	Synthesis methods	Via./cells	Ref.
a Note: Via. = viability; the cell viability value is picked for testing under GQD concentration of 100 μg mL^−1 and 24 h incubation time.
Top down	Photo-Fenton assisted cutting of GO	85%/MCF-7	78
	Oxidation cutting of graphite	76%/A549	77
	Oxidation cutting of graphite	95%/Hela, 93%/A549	73
	Microwave assisted cutting of GO	99%/Hela	68
	Hydrothermal cutting graphite in oxidizers	84%/THP1	85
	Electrochemical exfoliation of graphite rod	76%/U251	82
	Exfoliation of carbon fibre in mixed acids	78%/A549, 76%/KB	76
	Electrochemical unzipping MWCNTS		69
Bottom up	Pyrolysis of citric acid	92%/A549	75
	Pyrolysis of L-glutamic acid		80
	Alkali-mediated hydrothermal treatment of pyrene	61%/stem cells	83

---

Fig. 5 Cell viability values of A549 cells under the same testing conditions (i.e. concentration of GQDs = 100 μg mL⁻¹ and incubation time = 24 h) for GQDs from top-down and bottom-up methods in the literature.

4.3 Size vs. toxicity of GQDs

It has been known that the size of nano materials is a key factor dominating their toxicity.^{91,94,96–99} Indeed, it seems that smaller size is an advantage GQDs possess over GO in terms of toxicity. As compared in the work of Wu et al., the cytotoxicity of micron-meter-sized GO is lower than that of nano-meter-sized GQDs.⁷⁸ They suggested that the higher toxicity of GO results from the greater difficulty of internalization and higher ROS level generated than that of GQDs. Similar observations are also available from the work of Wang et al.⁷⁹ It is noticed that uptake of GO damaged the integrity of the cell membrane, causing hemolysis and aberrant forms, which did not happen during the uptake of GQDs. Apart from showing the different toxicity spectra of GO and GQDs via in vitro tests, in vivo tests with green gram sprouts also proved that GQDs have lower toxicity than GO or CNT.⁹⁰ Hence, the very fact that GO and CNT possess higher toxicity than that of GQDs suggests that size is a key factor for the low toxicity feature of GQDs. Although size can induce significant differences among the toxicities of GO, CNT and GQDs, how size differences among GQDs themselves influence their toxicity is still a vague question. Ignoring the types of cells being tested, we noticed that most of the GQDs that allow extremely high cell viability (i.e. over 90% for a concentration of 100 μg mL⁻¹ and incubation for 24 h) have size below 10 nm, as shown in Fig. 6. The cell viability for a size such as 20 nm is obviously lower than for many of the smaller ones. In the work of Mrakovic et al. showing the lethal effect induced by GQDs under blue irradiation to U251 cells, the size of the adopted GQDs was around 60 nm. These observations might imply that smaller GQDs are more biologically benign than larger ones. Nevertheless, we have not yet found a study which provides direct evidence for unveiling how the size of GQDs changes their toxicity profile, which is worth addressing in future in vitro and in vivo studies.

Fig. 6 Viability mapping for various cells under the same testing conditions (i.e. concentration of GQDs = 100 μg mL⁻¹ and incubation time = 24 h) from existing studies.

4.4 Surface chemistry vs. toxicity of GQDs

Surface chemistry in terms of the variety of functional groups on the surface of nano materials has been shown to be another highly important factor governing the toxicity of nano materials.^91,99 This is also true of GQDs. Firstly, we have shown in the previous section that different synthesis methods produce GQDs with divergent profiles of toxicity. One possible reason is the variety of surface chemistries from different synthesis routes. Moreover, Yuan et al. conducted an in vitro study with three types of GQDs which were defined by the different functional groups (i.e. –NH₂, –COOH and –CO–N(CH₃)) attached to the surface of the GQDs.⁷⁵ Although their experimental results suggest that all three GQDs have low cytotoxicity and excellent biocompatibility regardless of the applied chemical modification, one can note that the toxicity levels of the three GQDs are in fact not similar. Another recent report from Suzuki et al. showed that chiral GQDs as synthesized by two chiral moieties of L/D-cysteine have different biocompatibilities. It is learnt through both in vivo tests and dynamic simulation that L-GQD is more biocompatible than D-GQD.¹⁰⁰ In addition to the difference in functional groups, the percentage variance of functional groups may also lead to different toxicity traits. Unfortunately, we have not yet found any work on this topic. Lastly, the influence of surface chemistry on toxicity could also be indicated by the methods applied to regulate the toxicity of GQDs which will be discussed in a later section.

4.5 Doping vs. toxicity of GQDs

We have shown that the toxicity of GQDs is concentration dependant. Despite the common trend that toxicity increases with a rise in concentration of GQDs, test subjects present dissimilar extents of tolerance to the concentrations of different GQDs. With many of the raw GQDs having a cut-off concentration of 200 μg mL⁻¹ for cell viability above 80%, the doped GQDs can allow higher viability (e.g. over 90%) at a much higher concentration. For instance, the cell viability tests with boron-doped GQDs synthesized by Hai et al. showed that near 95% could be obtained at a concentration of 2 mg mL⁻¹ and even at a higher concentration of 4 mg mL⁻¹, the value is still 87% (ref. 68) (Fig. 7). Although there is still no direct evidence showing that nitrogen-doped GQDs share a similar effect with boron-doped GQDs, at least they did not render the toxicity and biocompatibility worse than raw GQDs.^72,82,101 These researches suggest that doping GQDs with benign foreign atoms such as boron and nitrogen is biologically safe.

Fig. 7 Cell viability of Hela cells incubated for 12 h (replotted with data from ref. 68).

5. Toxicity regulation

As the mechanism of toxicity such as ROS could not be completely excluded from raw GQDs, and in practice different applications may have varied requirements on toxicity level, it is likely that there are occasions when the toxicity of certain types of GQDs needs to be alleviated or driven further lower. Hence it is also necessary to develop ways to regulate the toxicity of GQDs. Previous research has revealed that the toxicity of nano materials (e.g. GO or semiconductor quantum dots) could be alleviated through various surface modification methods and moieties, including protein, RNA and polyethylene glycol (PEG),^{89,91,99,102,103} have been adopted. However, similar studies on GQDs are still rare. The only available researches were on regulating the toxicity of GQDs with PEG. Chong et al. who modified GQDs with PEG observed no difference in ROS level between GQDs-treated cells and the control group,⁷³ which suggests that the PEG-modified GQDs do not increase the ROS level of cells. Chandra et al. also synthesized PEG-modified GQDs and demonstrated that the biocompatibility of GQDs could be improved, as the PEG modification suppressed the ROS generation.⁶⁹ Hence, this is still an area needing further development.

6. Conclusion and future perspectives

In conclusion, we have summarized the recent research efforts on the toxicology of graphene quantum dots. In general, the majority of the existing studies suggest that GQDs have relatively low in vivo and in vitro toxicity and excellent biocompatibility, especially in comparison to their peers including graphene oxide (GO), carbon nanotubes and conventional semiconductor quantum dots. The low toxicity and excellent biocompatibility make GQDs a promising candidate for bio-applications such as bio-imaging, bio-sensing, and biomedicine (e.g. drug delivery). However, the toxicity profile of GQDs varied in the reports with different test subjects and different GQDs and there is also evidence suggesting that certain GQDs can cause the death of certain cells via the generation of intracellular reactive oxygen species (ROS). By going through the existing studies, we noticed several important aspects. Firstly, the toxicity of GQDs is highly determined by their concentration. Viability tests with various cells and certain in vivo tests (e.g. zebra fish embryos) indicate that the toxicity of GQDs increases with their concentration. Although high cell viability could even be obtained for some GQDs with a concentration as high as 0.5 mg L⁻¹, for many a ‘safe concentration’ cut-off is around 200 μg L⁻¹, and toxicity tends to be higher over this ‘safe concentration’. Secondly, the toxicity profile of GQDs varied with their synthesis methods. GQDs obtained from bottom-up methods might have better biocompatibility than those obtained through top-down methods. Thirdly, size is suggested to be the key advantage rendering the toxicity of nano-meter-sized GQDs lower than that of micron-meter-sized GO. In addition, it seems that large-sized GQDs have higher toxicity than small ones (<10 nm). Moreover, different surface chemistry states also lead to different toxicity features and GQDs doped with benign elements (e.g. boron) seem more biocompatible than raw GQDs. Lastly, a toxicity mechanism such as ROS could not be totally excluded for GQDs. Regulation of toxicity of GQDs is also attracting research effort. So far PEG has been applied to reduce the generation of ROS by GQDs.

Although the availability of toxicology studies of GQDs has been on the rise for the past few years, there is still an eminent need for more research effort to be spent in this field so that a full spectrum of the toxicity of GQDs could be revealed. Based on our evaluation of the current research status, we would like to suggest the following as possible future research perspectives.

(1) At the moment, more in vitro toxicology tests are available than in vivo. It is necessary to have more in vivo tests.

(2) More systematic studies should be carried out to decipher the correlation between the important parameters and the as-governed toxicity, such as methods of synthesis, particle size, surface chemistry and chemical compositions.

(3) By far the majority of the tests employed GQDs obtained from top-down methods. Continuous work on in vitro and in vivo toxicity tests with more GQDs obtained from bottom-up methods should be carried out.

(4) The majority of the current tests have shown that GQDs stay in the cytoplasm and no destructive effects have been observed on the nucleus and, therefore, GQDs should have a low risk for inducing genotoxicity. However, sufficient direct investigations are necessary to justify this postulation.

(5) Hitherto, the study of regulation of toxicity of GQDs is still rare (only PEG was adopted for this purpose). More efforts should be directed into this topic to search for ways to alleviate or further lower the toxicity of GQDs when needed.

(6) Last but not least, the testing conditions (e.g. incubation time and concentration) and evaluation criteria (e.g. the criteria for ‘high’ cell viability) in the existing research vary widely, making it difficult to cross-compare the toxicity of GQDs in different studies. Hence, given the variety of available GQDs, the establishment of standard testing and evaluation conditions might be necessary.

Acknowledgements

S. W. acknowledges the support of a Griffith International Postgraduate Scholarship (GIPS) and a CSIRO OCE top-up scholarship. Q. L. wishes to thank a Griffith University Research Infrastructure Grant. The authors thank the Centre for Microscopy and Microanalysis at the University of Queensland for the characterization facilities and support.

References

1. L. A. Ponomarenko, F. Schedin, M. I. Katsnelson, R. Yang, E. W. Hill, K. S. Novoselov and A. K. Geim, Science, 2008, 320, 356–358.
2. A. Barreiro, H. S. van der Zant and L. M. Vandersypen, Nano Lett., 2012, 12, 6096–6100.
3. K. Nakada, M. Fujita, G. Dresselhaus and M. S. Dresselhaus, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 17954–17961.
4. J. Guttinger, F. Molitor, C. Stampfer, S. Schnez, A. Jacobsen, S. Droscher, T. Ihn and K. Ensslin, Rep. Prog. Phys., 2012, 75, 126502.
5. C. Stampfer, S. Fringes, J. Güttinger, F. Molitor, C. Volk, B. Terrés, J. Dauber, S. Engels, S. Schnez, A. Jacobsen, S. Dröscher, T. Ihn and K. Ensslin, Front. Phys., 2011, 6, 271–293.
6. S. Fujii and T. Enoki, Acc. Chem. Res., 2013, 46, 2202–2210.
7. H. Q. Shi, A. S. Barnard and I. K. Snook, J. Mater. Chem., 2012, 22, 18119–18123.
8. B. Mandal, S. Sarkar and P. Sarkar, J. Nanopart. Res., 2012, 14, 1317.
9. M. Ezawa, Eur. Phys. J. B, 2009, 67, 543–549.
10. M. Ezawa, New J. Phys., 2009, 11, 095005.
11. J. Fernández-Rossier and J. Palacios, Phys. Rev. Lett., 2007, 99, 177204.
12. T. Espinosa-Ortega, I. A. Luk’yanchuk and Y. G. Rubo, Phys. Rev. B: Condens. Matter Mater. Phys., 2013, 87, 205434.
13. W. L. Wang, S. Meng and E. Kaxiras, Nano Lett., 2008, 8, 241–245.
14. L. A. Agapito, N. Kioussis and E. Kaxiras, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 82, 201411.
15. E. F. Sheka and L. A. Chernozatonskii, Int. J. Quantum Chem., 2010, 110, 1938–1946.
16. A. Güçlü and P. Hawrylak, Phys. Rev. B: Condens. Matter Mater. Phys., 2013, 87, 035425.
17. M. L. Mueller, X. Yan, J. A. McGuire and L. S. Li, Nano Lett., 2010, 10, 2679–2682.
18. Z. Z. Zhang and K. Chang, Phys. Rev. B: Condens. Matter Mater. Phys., 2008, 77, 235411.
19. S. Zhu, J. Zhang, S. Tang, C. Qiao, L. Wang, H. Wang, X. Liu, B. Li, Y. Li, W. Yu, X. Wang, H. Sun and B. Yang, Adv. Funct. Mater., 2012, 22, 4732–4740.
20. W. Chen, F. Li, C. Wu and T. Guo, Appl. Phys. Lett., 2014, 104, 063109.
21. Y. Wang, L. Zhang, R.-P. Liang, J.-M. Bai and J.-D. Qiu, Anal. Chem., 2013, 85, 9148–9155.
22. S. J. Wang, Z. G. Chen, I. Cole and Q. Li, Carbon, 2015, 82, 304–313.
23. H. Sun, N. Gao, L. Wu, J. Ren, W. Wei and X. Qu, Chemistry, 2013, 19, 13362–13368.
24. Y. Dong, W. Tian, S. Ren, R. Dai, Y. Chi and G. Chen, ACS Appl. Mater. Interfaces, 2014, 6, 1646–1651.
25. L. Zhou, J. Geng and B. Liu, Part. Part. Syst. Charact., 2013, 30, 1086–1092.
26. J. M. Bai, L. Zhang, R. P. Liang and J. D. Qiu, Chemistry, 2013, 19, 3822–3826.
27. F. Wang, Z. Gu, W. Lei, W. Wang, X. Xia and Q. Hao, Sens. Actuators, B, 2014, 190, 516–522.
28. Y. Dong, G. Li, N. Zhou, R. Wang, Y. Chi and G. Chen, Anal. Chem., 2012, 84, 8378–8382.
29. Y. H. Li, L. Zhang, J. Huang, R. P. Liang and J. D. Qiu, Chem. Commun., 2013, 49, 5180–5182.
30. Y. He, X. Wang, J. Sun, S. Jiao, H. Chen, F. Gao and L. Wang, Anal. Chim. Acta, 2014, 810, 71–78.
31. X. Wu, F. Tian, W. Wang, J. Chen, M. Wu and J. X. Zhao, J. Mater. Chem. C, 2013, 1, 4676–4684.
32. T. S. Sreeprasad, A. A. Rodriguez, J. Colston, A. Graham, E. Shishkin, V. Pallem and V. Berry, Nano Lett., 2013, 13, 1757–1763.
33. H. Huang, L. Liao, X. Xu, M. Zou, F. Liu and N. Li, Talanta, 2013, 117, 152–157.
34. Z. B. Qu, X. Zhou, L. Gu, R. Lan, D. Sun, D. Yu and G. Shi, Chem. Commun., 2013, 49, 9830–9832.
35. J. Ju and W. Chen, Anal. Chem., 2015, 87, 1903–1910.
36. J. Ju, R. Z. Zhang, S. J. He and W. Chen, RSC Adv., 2014, 4, 52583–52589.
37. J. Ju, R. Z. Zhang and W. Chen, Sens. Actuators, B, 2016, 228, 66–73.
38. X. Yan, X. Cui, B. Li and L. S. Li, Nano Lett., 2010, 10, 1869–1873.
39. Z. F. Wang, H. D. Zeng and L. Y. Sun, J. Mater. Chem. C, 2015, 3, 1157–1165.
40. J. Shen, Y. Zhu, X. Yang and C. Li, Chem. Commun., 2012, 48, 3686–3699.
41. M. Li, W. Ni, B. Kan, X. Wan, L. Zhang, Q. Zhang, G. Long, Y. Zuo and Y. Chen, Phys. Chem. Chem. Phys., 2013, 15, 18973–18978.
42. Y. Li, Y. Hu, Y. Zhao, G. Shi, L. Deng, Y. Hou and L. Qu, Adv. Mater., 2011, 23, 776–780.
43. S. J. Wang, I. Cole and Q. Li, Chem. Commun., 2016, 52, 9208.
44. S. Zhu, J. Zhang, C. Qiao, S. Tang, Y. Li, W. Yuan, B. Li, L. Tian, F. Liu, R. Hu, H. Gao, H. Wei, H. Zhang, H. Sun and B. Yang, Chem. Commun., 2011, 47, 6858.
45. X. Sun, Z. Liu, K. Welsher, J. T. Robinson, A. Goodwin, S. Zaric and H. Dai, Nano Res., 2008, 1, 203–212.
46. Z. H. Wang, J. F. Xia, C. F. Zhou, B. Via, Y. Z. Xia, F. F. Zhang, Y. H. Li, L. H. Xia and J. Tang, Colloids Surf., B, 2013, 112, 192–196.
47. T. Ihn, J. Guttinger, F. Molitor, S. Schnez, E. Schurtenberger, A. Jacobsen, S. Hellmuller, T. Frey, S. Droscher, C. Stampfer and K. Ensslin, Mater. Today, 2010, 13, 44–50.
48. L. Kou, F. Li, W. Chen and T. Guo, Org. Electron., 2013, 14, 1447–1451.
49. L. Zhao and S. F. Yelin, Phys. Rev. B: Condens. Matter Mater. Phys., 2010, 81, 115441.
50. D. Pan, J. Zhang, Z. Li and M. Wu, Adv. Mater., 2010, 22, 734–738.
51. D. Pan, L. Guo, J. Zhang, C. Xi, Q. Xue, H. Huang, J. Li, Z. Zhang, W. Yu, Z. Chen, Z. Li and M. Wu, J. Mater. Chem., 2012, 22, 3314.
52. L.-L. Li, J. Ji, R. Fei, C.-Z. Wang, Q. Lu, J.-R. Zhang, L.-P. Jiang and J.-J. Zhu, Adv. Funct. Mater., 2012, 22, 2971–2979.
53. F. Liu, M.-H. Jang, H. D. Ha, J.-H. Kim, Y.-H. Cho and T. S. Seo, Adv. Mater., 2013, 25, 3657–3662.
54. Y. Sun, S. Wang, C. Li, P. Luo, L. Tao, Y. Wei and G. Shi, Phys. Chem. Chem. Phys., 2013, 15, 9907–9913.
55. J. Peng, W. Gao, B. K. Gupta, Z. Liu, R. Romero-Aburto, L. Ge, L. Song, L. B. Alemany, X. Zhan, G. Gao, S. A. Vithayathil, B. A. Kaipparettu, A. A. Marti, T. Hayashi, J. J. Zhu and P. M. Ajayan, Nano Lett., 2012, 12, 844–849.
56. L. Minati, S. Torrengo, D. Maniglio, C. Migliaresi and G. Speranza, Mater. Chem. Phys., 2012, 137, 12–16.
57. Y. Q. Dong, H. C. Pang, S. Y. Ren, C. Q. Chen, Y. W. Chi and T. Yu, Carbon, 2013, 64, 245–251.
58. D. B. Shinde and V. K. Pillai, Chemistry, 2012, 18, 12522–12528.
59. S. J. Wang, I. S. Cole, D. Y. Zhao and Q. Li, Nanoscale, 2016, 8, 7449–7458.
60. L. B. Tang, R. B. Ji, X. K. Cao, J. Y. Lin, H. X. Jiang, X. M. Li, K. S. Teng, C. M. Luk, S. J. Zeng, J. H. Hao and S. P. Lau, ACS Nano, 2012, 6, 5102–5110.
61. S. J. Wang, Z. Lemon, I. S. Cole and Q. Li, RSC Adv., 2015, 5, 41248–41254.
62. J. H. Shen, Y. H. Zhu, X. L. Yang, J. Zong, J. M. Zhang and C. Z. Li, New J. Chem., 2012, 36, 97–101.
63. Y. Li, Y. Zhao, H. H. Cheng, Y. Hu, G. Q. Shi, L. M. Dai and L. T. Qu, J. Am. Chem. Soc., 2012, 134, 15–18.
64. X. B. Hou, Y. B. Li and C. Zhao, Aust. J. Chem., 2016, 69, 357–360.
65. S. Dey, A. Govindaraj, K. Biswas and C. N. R. Rao, Chem. Phys. Lett., 2014, 595, 203–208.
66. H. L. Fei, R. Q. Ye, G. L. Ye, Y. J. Gong, Z. W. Peng, X. J. Fan, E. L. G. Samuel, P. M. Ajayan and J. M. Tour, ACS Nano, 2014, 8, 10837–10843.
67. T. T. Zhang, H. M. Zhao, G. F. Fan, Y. X. Li, L. Li and X. Quan, Electrochim. Acta, 2016, 190, 1150–1158.
68. X. Hai, Q. X. Mao, W. J. Wang, X. F. Wang, X. W. Chen and J. H. Wang, J. Mater. Chem. B, 2015, 3, 9109–9114.
69. A. Chandra, S. Deshpande, D. B. Shinde, V. K. Pillai and N. Singh, ACS Macro Lett., 2014, 3, 1064–1068.
70. P. Roy, A. P. Periasamy, C. Y. Lin, G. M. Her, W. J. Chiu, C. L. Li, C. L. Shu, C. C. Huang, C. T. Liang and H. T. Chang, Nanoscale, 2015, 7, 2504–2510.
71. X. J. Wang, X. Sun, J. Lao, H. He, T. T. Cheng, M. Q. Wang, S. J. Wang and F. Huang, Colloids Surf., B, 2014, 122, 638–644.
72. R. Li, Y. S. Liu, Z. Q. Li, J. P. Shen, Y. T. Yang, X. D. Cui and G. C. Yang, Chem.–Eur. J., 2016, 22, 272–278.
73. Y. Chong, Y. F. Ma, H. Shen, X. L. Tu, X. Zhou, J. Y. Xu, J. W. Dai, S. J. Fan and Z. J. Zhang, Biomaterials, 2014, 35, 5041–5048.
74. D. Jiang, Y. P. Chen, N. Li, W. Li, Z. G. Wang, J. L. Zhu, H. Zhang, B. Liu and S. Xu, PLoS One, 2015, 10, e0144906,  DOI:10.1371/journal.pone.0144906.
75. X. C. Yuan, Z. M. Liu, Z. Y. Guo, Y. H. Ji, M. Jin and X. P. Wang, Nanoscale Res. Lett., 2014, 9, 108.
76. M. Nurunnabi, Z. Khatun, K. M. Huh, S. Y. Park, D. Y. Lee, K. J. Cho and Y. K. Lee, ACS Nano, 2013, 7, 6858–6867.
77. Y. Q. Sun, S. Q. Wang, C. Li, P. H. Luo, L. Tao, Y. Wei and G. Q. Shi, Phys. Chem. Chem. Phys., 2013, 15, 9907–9913.
78. C. Y. Wu, C. Wang, T. Han, X. J. Zhou, S. W. Guo and J. Y. Zhang, Adv. Healthcare Mater., 2013, 2, 1613–1619.
79. T. T. Wang, S. J. Zhu and X. E. Jiang, Toxicol. Res., 2015, 4, 885–894.
80. X. Wu, F. Tian, W. X. Wang, J. Chen, M. Wu and J. X. Zhao, J. Mater. Chem. C, 2013, 1, 4676–4684.
81. M. Zhang, L. L. Bai, W. H. Shang, W. J. Xie, H. Ma, Y. Y. Fu, D. C. Fang, H. Sun, L. Z. Fan, M. Han, C. M. Liu and S. H. Yang, J. Mater. Chem., 2012, 22, 7461–7467.
82. Z. M. Markovic, B. Z. Ristic, K. M. Arsikin, D. G. Klisic, L. M. Harhaji-Trajkovic, B. M. Todorovic-Markovic, D. P. Kepic, T. K. Kravic-Stevovic, S. P. Jovanovic, M. M. Milenkovic, D. D. Milivojevic, V. Z. Bumbasirevic, M. D. Dramicanin and V. S. Trajkovic, Biomaterials, 2012, 33, 7084–7092.
83. J. C. Qiu, D. S. Li, X. N. Mou, J. H. Li, W. B. Guo, S. Wang, X. Yu, B. J. Ma, S. Zhang, W. Tang, Y. H. Sang, P. R. Gil and H. Liu, Adv. Healthcare Mater., 2016, 5, 702–710.
84. J. Kim, S. H. Song, Y. Jin, H. J. Park, H. Yoon, S. Jeon and S. W. Cho, Nanoscale, 2016, 8, 8512–8519.
85. Y. R. Qin, Z. W. Zhou, S. T. Pan, Z. X. He, X. J. Zhang, J. X. Qiu, W. Duan, T. X. Yang and S. F. Zhou, Toxicology, 2015, 327, 62–76.
86. S. J. Zhu, J. H. Zhang, C. Y. Qiao, S. J. Tang, Y. F. Li, W. J. Yuan, B. Li, L. Tian, F. Liu, R. Hu, H. N. Gao, H. T. Wei, H. Zhang, H. C. Sun and B. Yang, Chem. Commun., 2011, 47, 6858–6860.
87. S. J. Zhu, N. Zhou, Z. Y. Hao, S. Maharjan, X. H. Zhao, Y. B. Song, B. Sun, K. Zhang, J. H. Zhang, H. C. Sun, L. J. Lu and B. Yang, RSC Adv., 2015, 5, 39399–39403.
88. Z. G. Wang, R. Zhou, D. Jiang, J. E. Song, Q. Xu, J. Si, Y. P. Chen, X. Zhou, L. Gan, J. Z. Li, H. Zhang and B. Liu, Biomed. Environ. Sci., 2015, 28, 341–351.
89. Y. L. Zhao, Q. L. Wu and D. Y. Wang, RSC Adv., 2015, 5, 92394–92405.
90. X. L. Li, Z. H. Zhou, D. J. Lu, X. W. Dong, M. H. Xu, L. M. Wei and Y. F. Zhang, Nanoscale Res. Lett., 2014, 9, 583.
91. H. Z. Sun, F. Zhang, H. T. Wei and B. Yang, J. Mater. Chem. B, 2013, 1, 6485–6494.
92. G. B. Qu, X. Y. Wang, Z. Wang, S. J. Liu and G. B. Jiang, Nanoscale Res. Lett., 2013, 8, 198.
93. H. J. Sun, N. Gao, K. Dong, J. S. Ren and X. G. Qu, ACS Nano, 2014, 8, 6202–6210.
94. S. H. Ouyang, X. G. Hu and Q. X. Zhou, ACS Appl. Mater. Interfaces, 2015, 7, 18104–18112.
95. W. H. Shang, X. Y. Zhang, M. Zhang, Z. T. Fan, Y. T. Sun, M. Han and L. Z. Fan, Nanoscale, 2014, 6, 5799–5806.
96. O. Akhavan and E. Ghaderi, ACS Nano, 2010, 4, 5731–5736.
97. K. H. Liao, Y. S. Lin, C. W. Macosko and C. L. Haynes, ACS Appl. Mater. Interfaces, 2011, 3, 2607–2615.
98. P. Zhang, Y. H. Ma, Z. Y. Zhang, X. He, Y. Y. Li, J. Zhang, L. R. Zheng and Y. L. Zhao, Nanotoxicology, 2015, 9, 1–8.
99. H. Kang, S. Mintri, A. V. Menon, H. Y. Lee, H. S. Choi and J. Kim, Nanoscale, 2015, 7, 18848–18862.
100. N. Suzuki, Y. C. Wang, P. Elvati, Z. B. Qu, K. Kim, S. Jiang, E. Baumeister, J. Lee, B. Yeom, J. H. Bahng, J. Lee, A. Violi and N. A. Kotov, ACS Nano, 2016, 10, 1744–1755.
101. Y. L. Zhao, Q. Liu, S. Shakoor, J. R. Gong and D. Y. Wang, Toxicol. Res., 2015, 4, 270–280.
102. W. B. Hu, C. Peng, M. Lv, X. M. Li, Y. J. Zhang, N. Chen, C. H. Fan and Q. Huang, ACS Nano, 2011, 5, 3693–3700.
103. A. X. Wang, K. F. Pu, B. Dong, Y. Liu, L. M. Zhang, Z. J. Zhang, W. Duan and Y. M. Zhu, J. Appl. Toxicol., 2013, 33, 1156–1164.

---