Introduction

Graphene has attracted significant interest in the areas of materials science, fundamental physics, and engineering since its discovery in 2004 (Novoselov et al. 2004, 2005). Owing to its unique two-dimensional structure and attractive physiochemical properties, graphene has been receiving exciting attention in many fields and primarily in those of biology and medicine. Graphene oxide (GO), a distinctive structure of sp2 carbon, oxygen, and hydrogen in variable ratios, can be obtained by oxidizing graphite in an acidic medium (Hummers and Offeman 1958; Marcano et al. 2010). The presence of abundant functional groups (epoxy, hydroxyl, carboxylic groups), conjugation system, large surface area, low cytotoxic effect, and low cost (Liu et al. 2008; Sun et al. 2008; Zhang et al. 2010a, c, 2011a; Akhavan and Ghaderi 2010; Yang et al. 2011a, 2012c; Lee et al. 2011; Chang et al. 2011; Bao et al. 2011; Liao et al. 2011; Xiao and Chen 2012; Wang et al. 2012; Akhavan et al. 2012; Tang and Cao 2012; Chng and Pumera 2013) have led to a surge of important potential in drug loading and delivery. The functional groups, decorated the basal planes and edges of GO layers, significantly revise the van der Waals interactions between graphene sheets (Bourlinos et al. 2003; Scholz and Boehm 2004) and impart desirable dispersability in water and several organic solvents (Titelman et al. 2005; Paredes et al. 2008; Moazzami Gudarzi 2012). Moreover, these functional groups allow GO conjugation with polymers and drugs through both covalent and non-covalent modification techniques (Kuila et al. 2012). Furthermore, the high surface area and π-conjugated structure of GO enabled it, as a platform, to immobilize with a number of substances, including metallic nanoparticles, drugs, and fluorescent molecules for different applications (Lu et al. 2009; Zhang et al. 2010a; Liu et al. 2010; Loh et al. 2010b; Bao et al. 2011; Jayakumar et al. 2012; Kim et al. 2012; Wate et al. 2012). As a consequence of this high chemical versatility, several research groups have recently explored functionalized GO as a nanocarrier for in vitro and in vivo drug delivery. The continuously increasing research effort for GO-based drug delivery in the last decade is shown in Fig. 1.

Fig. 1
figure 1

Number of published articles in the last decade pertaining to GO-based drug delivery applications (based on data taken from www.sciencedirect.com on April 15, 2013)

Full size image

The building of versatile drug carrier systems is considered a challenge on high capacity loading, efficient delivery, and specific targeting (Vashist et al. 2011; Liu et al. 2012; Szűts and Szabó-Révész 2012). Lui et al. focused on PEGylated nanographene oxide (NGO) for drug delivery of hydrophobic anticancer drug molecules via π–π stacking. Their results showed that NGO-functionalized polyethylene glycol (PEG) is biocompatible without evident toxicity (Liu et al. 2008). Zhang et al. synthesized GO-based folic acid-conjugated, in order to control co-loading of doxorubicin (DOX) and camptothecin (CPT). Their system shows specific cytotoxicity and targeting the breast cancer cell line Michigan Cancer Foundation-7 (MCF-7) (Zhang et al. 2010a). In addition, pristine graphene and GO have been investigated in vitro and in vivo in different cell types (Liao et al. 2011; Zhang et al. 2011b, 2012; Qin et al. 2012).

This review discusses and highlights the recent investigations of GO’s applications in the field of drug delivery. The structure and properties of GO, covalent and non-covalent functionalization, characterization techniques, cytotoxicity of GO, and cytotoxicity of functionalized GO are also discussed. Later, GO-based drug delivery systems based on in vitro cellular uptake and in vivo cancer therapy are presented. Finally, a comparative study between carbon nanomaterials, such as carbon nanotube (CNTs), nanodiamond (ND), fullerene (C60), and GO-based drug delivery systems is produced.

Graphene oxide

The structure and chemistry of GO have been discussed extensively in other reviews (Park and Ruoff 2009; Dreyer et al. 2010). The specific atomic structure of GO is not openly understood due to its random oxygen groups and irregular layer stacking. To this end substantial methods (models and experiments) have been developed toward understanding the GO structure. Hofman–Holst (2006), Ruess (Ruess 1946), and Nakajima–Matsuo (He et al. 1996) introduced model structures for GO. Lerf–Klinowski (Lerf et al. 1998) and Dekany (Buchsteiner et al. 2006) are the most well-known models using NMR to study the GO structure. Cai et al. (2008) used synthetic 13C-labeled graphite for GO preparation. Their results revealed that the Lerf–Klinowski and Dekany models as the preferable structures for GO. Figure 2 illustrates different model structures of GO, showing the oxygen groups localized at the edges of graphene sheet as well as above and below the basal plane.

Fig. 2
figure 2

Different model structures of GO: a Hofmann–Holst, b Ruess, c Nakajima–Matsuo, and d, e Lerf–Klinowski [deduced from Refs. (Lerf et al. 1998; Josepovits et al. 2006)]

Full size image

Structure and properties of GO

The chemical structure of graphene and GO is shown in Fig. 3. Over the years, considerable effort went toward understanding the structure of GO. In GO, the carbon atoms covalently bonded to oxygen containing groups are sp3 hybrids and can disrupt the sp2 conjugated system of graphene lattice structure (Gómez-Navarro et al. 2007; Eda et al. 2009). sp3 hybridized regions are randomly distributed either in or out of the basal plane of graphene sheets or at the edges (Schniepp et al. 2006).

Fig. 3
figure 3

Chemical structure of graphene and GO

Full size image

The properties of GO are unreliable and depend on its synthesis methods, processing, and experimental conditions. Electrically, GO is considered an insulating material due to its widespread saturated sp3 bonds, vacancies (missing carbon atoms), and negatively charged density species bound to carbon. For these reasons, the energy gap increases and makes GO non-conducting (Boukhvalov and Katsnelson 2008; Jung et al. 2008b; Yan and Chou 2010). Nevertheless, the structural and electronic properties of GO can be modified via chemical and thermal treatments (Gilje et al. 2007; Becerril et al. 2008; Eda et al. 2008). The optical properties also depend on the oxidation level of GO (Jung et al. 2007, 2008c). The multiple layers of GO change considerably the corresponding optical properties, for instance, the refractive index of thermally reduced GO is higher than that in stacked GO. GO is highly soluble and disperses well in water and physiological media, making it attractive for medical purposes after numerous modifications.

Recently, in order to delve into the GO structure in more detail, several microscopic and spectroscopic methods have been employed to investigate the structure of GO and its chemical compositions, as will be shown later. Such methods include atomic force microscopy (AFM), scanning transmission electron microscopy (STEM) combined with electron energy loss spectroscopy (EELS), scanning tunneling microscopy (STM), high-resolution transmission electron microscopy (HRTEM), Raman spectroscopy, Fourier transform infrared spectroscopy (FTIR), and X-ray photoelectron spectroscopy (XPS).

In addition, theoretical studies have been considered for GO structural exploration, thus providing significant insight in its workable kinetic and thermodynamic structure. Depending on the first-principle calculations, the building blocks in GO (atomic energy configurations) have been identified as containing epoxy and hydroxyl groups close to each other (Yan and Chou 2010; Wang et al. 2010b; Lu et al. 2011; Yuan et al. 2011). Different arrangements of the building blocks have been found to yield a local density approximation band gap in the range of a few electron volts. This principle implies the possibility of opening and tuning the band gap of the GO depending on the degree of oxidation (Yan and Chou 2010; Yuan et al. 2011). Density functional theory (DFT) calculations revealed that the epoxy group is formed between the oxygen atom and two adjacent carbon atoms on the graphene network, while the hydroxyl group is formed on the opposite side (Lahaye et al. 2009). Furthermore, theoretical calculations have been used to study the atomic structure of reduced GO as well as the chemical changes of oxygen containing groups during reduction processes (Paci et al. 2007; Bagri et al. 2010; Acik et al. 2011; Larciprete et al. 2011).

Preparation protocols

The most common chemical routes for GO preparation begin from expandable graphite by chemical exfoliation using various oxidizing agents such as potassium permanganate (KMnO4) and potassium chlorate (KClO3). These routes were comprehensively covered in previous reviews (Park and Ruoff 2009; Dreyer et al. 2010) and are summarized in Table 1. The first procedures for GO were developed by Brodie (1859), Staudenmeier (1898), and Hummers and Offeman (1958). Brodie (1859) treated graphite with KClO3 and fuming nitric acid (HNO3). Staudenmaier (1898) improved Brodie’s approach by slowly adding KClO3 over 1 week to a solution containing concentrated sulfuric acid (H2SO4), concentrated HNO3 (63 %), and graphite. The mass ratio of graphite to KClO3 was 1:10. The possibility of explosion and length of time required are the main drawbacks of this approach. Hummers and Offeman (1958) reported an alternative method, which is safer and also not time consuming. In this protocol, a water-free mixture of concentrated H2SO4, sodium nitrate (NaNO3), and KMnO4 are involved; the reaction was performed at 45 °C and continued for approximately 2 h. Minor modifications to the Hummers method were developed and still remain in use (Kovtyukhova et al. 1999; Hirata et al. 2004).

Table 1 The oxidation methods of graphite to graphite oxide
Full size table

Functionalization of graphene oxide

It is well known that carbon nanomaterials aggregate in cell culture media (buffers) caused by the charge screen effect. Therefore, surface modification is the key to render the solubility and the biocompatibility of carbon nanomaterials for biological systems. Depending on application purposes, two surface coating regimes are developed, including covalent and non-covalent approaches, allowing GO to be used in biological systems (Kitano et al. 2007; Bai et al. 2009; Veca et al. 2009; Yang et al. 2009c, 2012c; Choi et al. 2010; Widenkvist 2010; Englert et al. 2011). Before surface functionalization, the size distribution and individual separation of GO are essential for in vitro and in vivo drug delivery.

Covalent functionalization

The covalent functionalization of nanoscaled carbon nanomaterials (CNMs) is preceded by an oxidation of the graphite in acidic media with strong oxidizing agents, resulting in oxygen-rich groups. As described above, the GO is highly oxidized by oxygen groups with C/O ratio of 2:1. The presence of these functional groups allows various chemical routes already known in chemistry to functionalize GO (Fig. 4) (Loh et al. 2010a). The acylation reactions are the most common routes used for binding molecular moieties onto oxygen-rich groups at the edges of GO. Carboxylic acid groups at the edges have the ability to bind with octadecylamine to modify GO by long alkyl chains (reaction VI in Fig. 4) that can be tethered to amine-functionalized molecules (Xu et al. 2009). Depending on this covalent functionalization, NGO-functionalized PEG is used to obtain a biocompatible conjugation system which can be performed as a platform for drug delivery (Sun et al. 2008). It is known that aromatic drug molecule 7-ethyl-10-hydroxycamptothecin (SN38) is water insoluble and hard to be used for the treatment of diseases. The resulting NGO–PEG/SN38 conjugation exhibits high water solubility while retaining a high cancer cell therapy similar to that of the free drug in organic solvents. The water soluble SN38 prodrug has been investigated in colon cancer treatment (Liu et al. 2008). The epoxy groups can undergo nucleophilic ring-opening reaction with amine-terminated molecules (reaction VIII in Fig. 4). For instance, octadecylamine (Wang et al. 2008), 1-(3-aminopropyl)-3-methylimidazolium bromide (Yang et al. 2009c), and 3-aminopropyltriethoxysilane (Lin et al. 2011; Wang et al. 2011) have been reacted with epoxy groups on the GO surface and afforded colloidal suspensions of GO in various polar solvents, such as water, DMF, and DMSO (Wang et al. 2008; Sun et al. 2010). The resultant amine-functionalized GO has been used in various applications, such as optoelectronic (Xu et al. 2009), biodevices (Mohanty and Berry 2008), polymer composite (Veca et al. 2009), and drug delivery (Liu et al. 2008).

Fig. 4
figure 4

Schematic illustration showing various covalent functionalizations of graphene or GO: I Reduction of GO [1 NaBH4, 2 KOH/H2O, 3 N2H4]. II Covalent functionalization of reduced graphene via diazonium reaction (Si and Samulski 2008; Lomeda et al. 2008). III Functionalization of GO by the sodium azide. IV Reduction of azide-functionalized GO with LiAlH4 producing amino-functionalized GO. V Functionalization of azide–GO through (R–ChCH/CuSO4). VI Modification of GO with the acylation reaction. VII Esterification of GO by activation of the COOH groups (1 DCC/DMAP or SOCl2, 2 ROH). VIII Nucleophilic ring-opening reaction. IX Covalent functionalization of graphene oxide via organic isocyanates. Reproduced with permission from Marques et al. (2011)

Full size image

Non-covalent functionalization

Non-covalent functionalization with different chemical substances is essential and considered to have less impact on the structure and properties of graphene. The attachment of functional groups to GO occurs via π–π, electrostatic binding, H-bonding, van der Waals, H–π, cation–π, and anion–π interactions (Tarakeshwar et al. 2001; Grimme 2004; Yi et al. 2006, 2009; Lee et al. 2007; Singh et al. 2009; Riley et al. 2010; Ma and Dougherty 2012). The π–π interaction is one of the most interesting non-covalent interactions; the diffuse electron clouds in the π systems exhibit attractive forces. For sufficient stability in aqueous solutions, graphene and GO via non-covalent interaction usually occur with surfactant molecules or amphiphilic polymers. However, biocompatible polymers for RGO surface coating are more useful than small surfactant molecules (Park et al. 2010). For the first time, via non-covalent functionalization, PEGylated NGO were employed as a nanocarrier to load therapeutic anticancer drugs and its cellular uptake was studied (Liu et al. 2008; Sun et al. 2008). Hu et al. developed graphene–pluronic F127 (PF 127) nanohybrid, the hydrophobic moieties of PF 127 attached to graphene surface via hydrophobic binding and the hydrophilic chains of PF 127 remained free to move in solution (Hu et al. 2012). PEG-grafted poly(maleic anhydride-alt-1-octadecene) via non-covalent produced sufficient physiological stability and increases blood circulation half-life in photothermal cancer therapy (Yang et al. 2012a, b). GO sheets possessed high negative charge; positively charged molecules could bind via electrostatic interactions. Liu et al. used polyethyleneimine (PEI) as cationic polymer to non-covalent coat GO and the results showed that the obtained GO-PEI has improved stability in culture media, high gene transfection efficiency, as well as reduced toxicity against treated cells (Feng et al. 2011). Tan et al. introduced l-proline/GO hybrid through hydrogen bonding interaction; the result revealed that the loading of l-proline onto GO (Fig. 7d) is highly efficient in hybrid catalysts compared with unloaded l-proline for the direct asymmetric aldol reaction (Tan et al. 2013a).

In a nanocarrier system, Depan et al. used positively charged polymer to encapsulate DOX-loaded GO, obtaining DOX–GO–chitosan–folate nanocarrier system that exhibited a pH response drug release profile (Depan et al. 2011). Protein-coated GO, demonstrated in many reports (Liu et al. 2011; Hu et al. 2011; Shen et al. 2012; Tan et al. 2013b), sonication of fetal bovine serum non-covalent-functionalized GO exhibits low cytotoxicity compared with uncoated GO (Hu et al. 2011). Owing to their high conjugation system (π–π delocalization regime), pristine graphene and GO possess the ability to bind with aromatic compounds through π–π stacking, including cancer drugs, fluorescence molecule (fluorescein), catalytic hybrid, and combinations of molecules.

Characterization techniques

A variety of characterization techniques have been used to utilize the structure and properties of GO. These techniques are classified into spectroscopic and microscopic approaches. The spectroscopic approaches are used to identify the chemical structure of GO, and include Raman, FTIR, and XPS. Microscopic tools are used to map out the structure of GO at various heights and lateral dimensions. For instance, AFM, SEM, TEM, and STM (Kovtyukhova et al. 1999; Sun et al. 2008; 2010; Lomeda et al. 2008; Loh et al. 2010b; Englert et al. 2011; Han et al. 2011; Du et al. 2011; Chang et al. 2011; Xiao and Chen 2012; Cheng et al. 2013).

Spectroscopic approaches

Raman spectroscopy

Raman spectroscopy, introduced by Krishna and Raman in 1981 (Rousseau et al. 1981), is a spectroscopic technique that provides information about molecular vibrations and can be used for sample identification and quantification. It is based on Raman scattering when monochromatic light hits a sample. The Raman effect occurs when light impinges upon a molecule and interacts with the electrons that bond to that molecule. As a result, a photon excites the molecule from the ground state to a virtual energy state. Later, the molecule returns back to a different rotational or vibrational state emitting a photon in the process. A detector detects the energy of the released photon which equals to the energy difference between the original state and the new molecule state. This energy difference leads to an up or down shift in the emitted photon’s frequency away from the excitation wavelength frequency. Raman shifts can provide information about vibrational, rotational, and other low frequency changes in the samples (Dresselhaus et al. 2004; Ferrari and Robertson 2004; Reich and Thomsen 2004).

Raman scattering is a powerful characterization technique used as a probe of order and disorder in the carbon skeleton of sp2 and sp3 hybridized carbon-based materials. Graphite material (multiple graphene sheets) has three dominant Raman features at ~1,580 cm−1 (G bank), ~1,350 cm−1 (D band), and ~2,700 cm−1 (2D band) as shown in a of Fig. 5. The G mode corresponds to planar vibrations of carbon atoms and is present in most graphite-like materials. The D mode is related to the structural defects and is present in all graphite-like carbon materials. A weak band at ~3,248 cm−1 (2D′ band) corresponds to an overtone of D′ (1,620 cm−1) mode (Das et al. 2008). One layer of highly crystalline graphene sheet has a 2D mode at lower frequency in comparison with graphite, the width of this band being around 24 cm−1 for monolayer graphene and varies for graphite from 45 to 60 cm−1. Moreover, the relative intensity of 2D and G bands—I 2D/I G in monolayer graphene sheet is higher than in graphite (Pimenta et al. 2007; Cançado et al. 2008; Malard et al. 2009). In GO, the G band is wider and shifted to a lower frequency region ca. 1,590 cm−1, the D mode intensity increases and probably becomes higher than the G mode due to the structural disorder in the sp2 pattern induced by oxygen containing groups on the carbon basal plane or at the edges. On the other hand, the 2D mode reduces and becomes wider with respect to the D and G bands (Kudin et al. 2008; Yang et al. 2009a; Wilson et al. 2009). Typical Raman spectra of GO are shown in Fig. 5a. Raman spectroscopy bands D, G, and 2D might give insight after graphene or GO functionalization; Zhong et al. functionalized graphene with Aryne molecules, the results showing the I D/I G ratio of the functionalized graphene increasing for all Arynes after covalent functionalization (Zhong et al. 2010). Further functionalization of GO shows no significant changes when characterizing with Raman spectroscopy (Pasricha et al. 2009; Zhang and Zhang 2011).

Fig. 5
figure 5

Typical spectra for the GO as gathered by a Raman spectroscopy (Makharza et al. 2013), b FTIR (Marques et al. 2011), and c XPS (Xue et al. 2012)

Full size image

Fourier transform infrared spectroscopy (FTIR)

Another spectroscopy technique yielding complementary information to those provided by Raman spectroscopy is infrared spectroscopy (IR). It excites a sample under observation with light in the infrared region of the electromagnetic spectrum. The infrared portion of the electromagnetic spectrum is divided into three regions: the near-, the mid- and the far-infrared. The mid-infrared is usually used to study the fundamental vibrations and associated rotational–vibrational structure, and the most related to the study of GO, as it contains the most useful vibrational frequencies of various oxygen groups. IR is based on the fact that the molecules absorb specific frequencies which match the frequency of the bond or group that vibrates. Radiation in the IR range is passed through the sample under investigation while the absorption coefficient of each wavelength is measured. The absorption of the radiation depends on the vibrational modes in the sample structure. Hence, each structure, depending on the existing vibrational modes, has its characteristic absorption spectrum (fingerprints). Consequently, the IR spectrum is frequently used in order to identify the presence or absence of specific functional groups in a molecule. In the case of GO, the most characteristic peaks are the broad peak of hydroxyl group (–OH) at 3,430 cm−1, the peak at 1,720 cm−1 corresponding to carbonyl group (C=O), 1,570 cm−1 representing the skeletal vibrations of sp2 carbon atoms, 1,225 cm−1 and ca. 1,100 cm−1 referring to stretching vibrations of COOH and C–O–C, respectively (Fig. 5b) (Stankovich et al. 2006; Xu et al. 2009; Yang et al. 2011b; Acik et al. 2011).

X-ray photoelectron spectroscopy (XPS)

XPS, also called electron spectroscopy for chemical analysis (ESCA), is a quantitative spectroscopic technique that utilizes photo-ionization and analysis of the kinetic energy distribution of the emitted photoelectrons in order to study the elemental composition, empirical formula, chemical state, and electronic state of the elements that exist within a material. XPS spectra are created by irradiating a material with a beam of X-rays in ultra-high vacuum (UHV) conditions, forcing the core electrons to be excited into unoccupied atomic/molecular orbitals above the Fermi level. Simultaneously, the kinetic energy in addition to the number of electrons that escape from the top few nanometer of the material being analyzed is measured.

This technique is particularly useful for identification of carbon nanomaterials, in which the chemical composition greatly influences their properties. High-resolution XPS survey scan provides information not only about the identity of elements, but also the ratios of these elements detected. In addition, it provides insight into the identity of the functional groups distributed onto the surface which modulate the carbon nanomaterial’s properties. For instance, XPS spectra of graphite have a strong peak, corresponding to C–C sp2 at 284.5 eV, and small one at 289 eV from plasmons (Briggs and Seah 1990; Vickerman 1997), which represent collective behavior of the delocalized electrons. In GO, XPS further unambiguously exhibits the carbon and oxygen bonds in their various forms; C–C (sp2 or sp3), C–OH, C–O–C, C=O, and C=OOH. The C1’s signal of these functional groups reveals at 284.5, 285.8, 286.5, 287.5, and 289.2 eV, respectively (Yang et al. 2009a; Mattevi et al. 2009; Lee et al. 2009; Akhavan 2010; Ganguly et al. 2011), as shown in c of Fig. 5. O1s provide complementary information to those in the C1s spectra, the deconvolution of O1s produces three main peaks around 531.1, 532.4, and 533.4 eV which are ascribed to C=O (oxygen atom connected to aromatic carbon), C–O (oxygen atom connected to aliphatic carbon), and C–OH (oxygen atom connected to aromatic carbon to form phenolic groups), respectively (Schniepp et al. 2006; Stankovich et al. 2007; Mattevi et al. 2009; Bagri et al. 2010). Intercalated adsorbed water molecules appeared at higher binding energy (534.7 eV) (Akhavan 2010). Moreover, XPS provides information about the identity of different kinds of metal ions like K, Na, Cu, Ni, Co, Cu, Ag, Mg, Pt, etc., as well as information on the valences and the ratio of these metals (Marcus and Maurice 2006; Stankovich et al. 2007; Altavilla and Ciliberto 2010).

Microscopic approaches

Microscopes have been important instruments for nano-structured materials. Various techniques are being used to understand the surface features of nanomaterials, such as optical microscopy, AFM, scanning electron microscopy, and high-resolution transmission electron spectroscopy.

Optical microscopy

For general evaluation of graphene and GO sheets, optical microscopy remains essential, a high throughput, and low cost imaging tool. Relatively, single-graphene sheet absorbs 2.3 % of the visible light (Nair et al. 2008; Duong et al. 2012). Reflected light microscope can provide high contrast images of graphene and GO (Novoselov et al. 2005). This allows researchers to scan the surfaces rapidly and measure the distribution and the shape of the GO flakes (Novoselov et al. 2004), as shown by the typical representative optical image in Fig. 6a.

Fig. 6
figure 6

Representative microscopy characterization of GO with a optical microscopy (Zhao et al. 2010), b AFM (Zhang et al. 2011c), and c SEM (Zhao et al. 2010)

Full size image

Atomic force microscopy (AFM)

AFM is a high-resolution type of scanning probe microscopy. It is a method to visualize the surface topology of materials in three-dimensional (3D) detail down to the nanometer scale (Haugstad 2012). AFM reveals the thickness of graphene and GO sheets, as well as the number of layers (Stankovich et al. 2007; Gómez-Navarro et al. 2007; Cote et al. 2009; Paredes et al. 2009; Cheng et al. 2013). Moreover, it can show the distance between two graphene or GO layers which is essential to distinguish between them. The topological information of the surface is gathered by “feeling” the surface with a mechanical probe (tip) fixed on a cantilever. This mechanical probe (tip) scans the measured surface in a raster scan manner, while the van der Waals forces determine the deflection of the cantilever. Piezoelectric elements that facilitate tiny, accurate, and precise movements control the probe motion. AFM uses a laser beam detection system in order to control the piezoelectric element movements, where the laser is reflected from the back of the deflected cantilever onto a position-sensitive detector. AFM works in three main modes: tapping mode, contact mode, and non-contact mode AFM (Warner et al. 2012). Tapping mode is frequently used in characterizing GO as it allows high-resolution imaging without inducing destructive frictional forces onto the sample under investigation (Zhong et al. 1993).

In GO, due to the presence of oxygen groups, the interlayer distance between GO sheets increased to 0.7 nm, which is roughly twice the distance found in bilayer graphene sheets (Fig. 6b). Thus, the expected height measurements by AFM might exceed that of graphene. Shen et al. (2009) investigated the thickness of individual GO and reduced GO nanoplatelets. The results showed the average thickness of an exfoliated GO nanoplatelets is ca. 1.3 nm while the reduced GO sheets exhibited a bumpy structure with flat areas showing heights of 0.2–0.4 nm and some high areas manifested an average height of 1.5 nm. The bumpy regions attributed to the dead space due to vast edge functionalization (Dideykin et al. 2011). GO-based polymer functionalization showed an increase in the height more than that of pristine GO (Layek et al. 2010). The average thickness of graphene-functionalized polymer (methyl methacrylate—MMA) is 3 nm (Fang et al. 2009). Two different sites of functionalization (edge and graphene basal planes) are mostly responsible for height distributions. NGO–PEG/DOX drug loading produced an obvious increase in thickness compared with the precursor NGO (Sun et al. 2008). The same results deduced after loading of CPT and SN38 onto NGO–PEG as a nanocarrier for drug delivery systems (Liu et al. 2008).

Scanning electron microscopy (SEM)

SEM is one of the most frequently used techniques in sample characterization, due to its good resolution, ease of applicability, large depth of focus, and high magnification. It is a type of electron microscope which images a sample by raster scanning it with a focussed beam of high-energy electrons. In a typical setup, an electron gun emits a high-energy electron beam in a high vacuum chamber, which later passes through a series of focussing and accelerating magnetic lenses. The high kinetic energy, carried by the accelerated electrons, is dissipated as a variety of signals produced by electron–sample interactions when the incident electrons are decelerated in the solid sample, including, in addition to many, the secondary electrons. The generated signals are then drawn to the secondary electron detector which is highly positively charged and guided through the Faraday cage to the collection target. Finally, they are converted into 2D grayscale images.

SEM is a versatile method used to gain information on the graphene and GO domains like size, shape, and nucleation density (Fig. 6c). It has also, recently, been used to study monolayer graphene onto different surfaces (Kim et al. 2009; Wood et al. 2011; Takahashi et al. 2012).

Cytotoxicity of graphene oxide

Graphene paper has been developed to be a biocompatible platform for adhesion and proliferation of murine fibroblast cells (L-929) (Chen et al. 2008), neuroendocrine cells (PC12), oligodendroglia cells (OLC), and osteoblasts (OBS) (Agarwal et al. 2010). Recently, in vivo cancer treatment using graphene as nanocarrier has been performed in animal experiments (Yang et al. 2010, 2011a). Various cytotoxicity investigations have been evaluated on graphene, GO, and functionalized GO in different cell lines (Zhang et al. 2010c, 2012; Wang et al. 2010a; Chang et al. 2011; Hu et al. 2011; Liao et al. 2011; Yang et al. 2012c). GO is a highly biocompatible material, thus, it inspirits the proliferation and adhesion of kidney cells, OBS, and human embryonic stem cells (hESc) (Agarwal et al. 2010; Park et al. 2010). On the contrary, other studies showed GO nanosheets mixed to cell culture media at 20 μg/mL can produce 20 % reduction in cell viability, whereas 50 μg/mL GO induce 50 % loss in cell viability. This result was attributed to inhibitory effect due to the GO suspension (Hu et al. 2010). Wang et al. examined human fibroblast cells (hFBc) with GO at different concentrations and found that concentration dependent on cytotoxicity when the concentration exceed 50 μg/mL (Wang et al. 2010a). For the first time, the cyto- and geno-toxic effect of reduced GO nanoparticles were investigated with mesenchymal stem cells (hMSCs) (Akhavan et al. 2012), with the study showing that the size and the concentration of GO are effective parameters on the hMSCs cytotoxicity.

Among these contradictory findings, there is no consensus on the cellular toxicity of GO, with different studies showing different biocompatibility of GO. For this inconsistency, Pumera and Chng reported the toxicity of GO depending on the oxidative methods used. They prepared GO by four different preparation methods (Staudenmaier—ST, Hofmann—HO, Hummers—HU, and Tour—TO), and investigated GO in adherent lung epithelial cells (LEc) by using in vitro MTT and WST-8 viability assays. The results revealed that different oxidizing methods exhibited different toxicological behavior of GOs as they contain different oxygen-based groups. All four GO nanomaterials’, GO-ST, GO-HO, GO-HU, and GO-TO, exposure with adenocarcinomic human alveolar epithelial cells (A549) media are dose-dependent cytotoxic responses. Varieties of oxygen groups and (C/O ratio) play an important role in the toxicity of GO nanoplatelets (Chng and Pumera 2013). It is generally agreed that the oxygen groups, preparation methods of GO, size, charge, and the structural defects of graphene might disturb its in vivo and in vitro behavior and its toxicity in biological applications.

Cytotoxicity of functionalized GO

As previously discussed, GO needs to be functionalized in order to perfectly disperse in culture media, as well as being compatible with tissue organisms. Indeed, GO revealed two types of functionalization (covalent and/or non-covalent) with small molecules or macromolecules, depending on the chemical structure of materials used for GO modification. The toxicity of pristine graphene and GO to mice are dose-dependent (Wang et al. 2010a; Yan et al. 2011; Zhang et al. 2011c). Functionalized NGO with biocompatible materials such as PEG (Sun et al. 2008; Yang et al. 2011a), dextran (Zhang et al. 2011a), chitosan (Fan et al. 2010; Liao et al. 2011), pluronic (Duch et al. 2011), tween (Park et al. 2010), low generation polyamidoamide (PAMAM) (G0) dendrimer (reference—our previous work), polyvinylpyrrolidone (PVP) (Qin et al. 2012), gelatin (An et al. 2013), sulfonic acid groups (Zhang et al. 2010a), and protein (fetal bovine serum—FBS) (Hu et al. 2011), show high reduction in vitro and in vivo toxicity. In the same context, Zhang et al. measured the cell viability percent of NGO–PEG vehicles as a function of concentration (Zhang et al. 2011b), and the data showed that above 95 % of mouse breast cancer cell line (EMT6) remained viable even at high concentration up to 100 μg/mL, which revealed that NGO–PEG has no cytotoxic effect to EMT6 cells after 24 h incubation. In other studies, gelatin-functionalized graphene nanosheets exhibited very low cytotoxic effect against A549, even at high concentration (300 μg/mL).

Surface modification of GO with appropriate biocompatible materials increases the uptake impact on various cell lines and retorts on the negative performance of pristine graphene and GO via a number of mechanisms such as relative oxygen species (ROS) and cell wall membrane damage.

Loading and delivery of drugs

In vitro tests

The large surface area of graphene makes it a strong candidate for drug loading other than commonly used carbon nanomaterials. It has a significant capability to interact with aromatic anticancer drugs and water insoluble drug molecules via strong π–π interaction (Sun et al. 2008; Liu et al. 2008). Doxorubicin (DOX) and water insoluble SN38 are the first two drug molecules loaded on PEGylated (PEG) nanographene oxide (NGO–PEG) (Fig. 7a). DOX loading onto NGO–PEG achieved by simple mixing of 0.5 mmol/L with NGO–PEG solution (~0.2 mg/mL) at pH 8 overnight via physisorption π stacking. NGO–PEG–Rituxan/DOX with Raji B—cells investigated in vitro (Fig. 7b), the result revealed that DOX delivery into Raji B—cells enhanced in NGO–PEG–Rituxan/DOX comparing with free DOX, NGO–PEG/DOX. Thiolated Rituxan (CD20 + antibody) conjugated to the amine groups onto NGO–PEG is used for selective killing of cancer cells (Sun et al. 2008). Also in this study, approximately 40 % of DOX released after 24 h in acidic solution (pH 5.5), this value of pH ascribed to increase the solubility and hydrophilicity of DOX for drug delivery analog, as well as accelerate the drug releasing (Fig. 7c). Zhang et al. found over 68 % of DOX released after 7 days at pH 5.5 (Zhang et al. 2011b). The pH of the medium is crucial in controlled drug delivery applications, for instance, extracellular tissues of cancer cells. In addition to the intracellular lysosomes and endosomes are acidic (Gillies and Fréchet 2005), which will drive the drug to be released from NGO–PEG structure. The loading quantity of DOX onto NGO–PEG was calculated to be 142.5 wt%, this amount of loading is higher than that in common drug delivery materials (Murakami et al. 2004; Liu et al. 2007b) which is always less than 100 %.

Fig. 7
figure 7

a A schematic illustration of a 6-arm branched PEG covalently bound to GO and SN 38 loading onto NGO–PEG via π–π stacking. b Relative cell viability of free and conjugated DOX at 2 and 10 μM concentrations via MTS assay. c Percentage of DOX retained versus time onto NGO–PEG in PBS solution at pH 5.5 and 7.4 [a, b, c were deduced from Ref. (Liu et al. 2008; Sun et al. 2008)]. d Non-covalent H-bonding of l-proline/GO hybrid (Tan et al. 2013a)

Full size image

The drug model (methotrexate—MTX) loaded on gelatin–GNP showed that pH-dependent release at low pH was better than in neutral conditions (An et al. 2013). MTX@gelatin–GNP showed lower cytotoxic effect with A549 cells in comparison with free MTX at the same concentration. Water insoluble molecule SN38 loaded NGO–PEG via π–π stacking revealed excellent solubility in the biological environment (Liu et al. 2008), NGO–PEG/SN38 exhibited high influence to kill cancer cells in vitro, with a human colon cancer cell line (HCT-16). Folic acid-loaded NGO with the two anticancer therapeutic drugs (DOX and CPT-11) showed specific targeting to MCF-7 human breast cancer cells, as well as high cytotoxicity compared to unmodified NGO loaded with DOX or CPT.

In vivo tests

The in vivo therapeutic efficacy of NGO–PEG/DOX was demonstrated through combination of photothermal treatment and chemotherapy (Zhang et al. 2011b). As shown in Fig. 8a, four mice groups were treated with PBS (200 μL), DOX (10 mg/kg, 200 μL), NGO–PEG (7 mg/kg, 200 μL), and NGO–PEG/DOX (10 mg/kg, 200 μL). The tumor region of NGO–PEG and NGO–PEG/DOX irradiated by NIR light (2 W/cm2, 5 min) after 24 h of injection.

Fig. 8
figure 8

a Photographs of mice groups after various treatments. b Tumor growth curves of different mice groups after treatment. c Body weights of mice groups after treatment. Deduced from Ref. (Zhang et al. 2011b)

Full size image

The time dependence of tumor volume is presented in Fig. 8b (the tumor volume normalized with respect to its initial size). Free DOX exposed rapid growth of tumor volume as a result of insufficient dosage to reduce it. NGO–PEG group showed reduction in the tumor volume after a few days of injection, and then follows its growth to reach the size of DOX group. NGO–PEG/DOX showed considerable trend with size reduction of tumor volume along a period of 30 days, this group demonstrated as a powerful vehicle for combined chemo-photothermal therapy of cancer in vivo. Figure 8c showed the weight loss of mice for the four groups as a function of time during the treatment. The result revealed that no weight loss was perceived, indicating that the toxicity of materials was not severed. On the other hand, Yang et al. reported for the first time the behavior of nanographene sheets in vivo by using PEG coating via a fluorescent labeling method, PEG-functionalized GO was labeled with the fluorescent dye Cy7 for in vivo investigations, the majority of Cy7 dye molecules are covalently bonded to NGO–PEG via an amide bond formation. Three tumor models (4T1 tumor bearing Balb/c mice, KB, and U87MG tumor bearing nude mice) were demonstrated with different time points after intravenously injected by NGO–PEG–Cy7 (200 μL of 2 mg/mL solution for each mouse; a dose of 20 mg/kg) and subsequently imaged.

The concentration of NGO–PEG–Cy7 versus time was measured and the blood circulation half-life was observed approximately at 1.5 h (Fig. 9a). As shown in Fig. 9b, NGO–PEG–Cy7 existed in different tumor models over time, NGO–PEG exhibiting an excellent in vivo tumor near-infrared (NIR) photothermal therapy without obvious toxicity to the treated mice. Neither death nor noteworthy body weight drop was observed in the NGO–PEG plus laser-treated materials, as shown in c of Fig. 9. After 40 days of photothermal therapy, the major organs of treated mice were collected for histology analysis. The results exhibited no noticeable signal of organ distortion (Fig. 9d), suggesting the assure of using PEGylated NGS for in vivo applications (Yang et al. 2010). Table 2 summarizes the cytotoxicity and the cellular uptake in vitro and in vivo for various therapeutic systems.

Fig. 9
figure 9

a The blood circulation profile of NGS–PEG–Cy7 measured by Cy7 fluorescence in the blood at different time points post-injection. b In vivo fluorescence images of mice models at different time points post-injection of NGS–PEG–Cy7. Mouse autofluorescence was subtracted and high tumor uptake of NGS–PEG–Cy7 was observed for all the three tumor models. Hairs on Balb/c mice were removed before fluorescence imaging. c Body weight curves after diverse treatments. d H&E stained images of major organs. No noticeable change was observed in major organs including kidney, liver, spleen, heart, intestine, and lung [reproduced with permission from Ref. (Yang et al. 2010)]

Full size image
Table 2 In vitro and in vivo toxicity evaluation of therapeutic nanomaterial systems
Full size table

Cytotoxicity of other carbon-based nanomaterials

Several studies emphasized that numerous factors such as chemical composition, size, shape, contaminants, concentration, and cell types will influence the cellular uptake and the cytotoxicity of carbon-based materials (Lam et al. 2004; Warheit et al. 2004; Sayes et al. 2005; Tsai et al. 2006; Schrand et al. 2007b; Liu et al. 2008; Casey et al. 2008; Shinohara et al. 2009; Yuan et al. 2010; Arlt et al. 2010; Zhang et al. 2010b; Chang et al. 2011; Liao et al. 2011; Yang et al. 2013; Chng and Pumera 2013). To the best of our knowledge, there are few experimental reports that compare the cytotoxicity of carbon nanomaterials (Schrand et al. 2007a; Zhang et al. 2012). Table 3 highlights the cytotoxicity evaluation of carbon nanomaterials concerning the size, concentration, and biological systems. Compared with CNTs, PEGylated NGO exhibited distinctive in vivo behaviors such as reduced reticuloendothelial systems (RES) accumulation and particularly improved tumor passive targeting effect (Liu et al. 2007a). The unique 2D structure, small size (10–50 nm), and biocompatibility played an important role to enhance permeability and the retention effect of NGS for high tumor passive uptake. Compared with nanoparticles, such as gold nanorods (AuNRs) significantly explored as photothermal agents, the result showed that PEGylated NGS emerged as comparable with PEGylated AuNRs in terms of administration routes (intravenous), injected doses (20 mg/kg), NIR laser densities (2 W/cm2), and irradiation durations (5 min) (von Maltzahn et al. 2009).

Table 3 Toxicity evaluation of carbon nanomaterials
Full size table

Fullerene is the first carbon nanomaterial investigated toward biomedical applications due to it being the first to be discovered, in 1985 (Krotto et al. 1985), and classified as an inorganic nanoparticle with wide availability due to its small size (~1 nm). Innate fullerene particles have very low dispersibility in water and form negative charge aggregates with an average size of 160 nm. Hence, OH groups and other organic molecules are induced in fullerene in order to activate its surface for better dispersibility in water and physiological media (Bosi et al. 2004; Han and Karim 2008) high dosages of fullerene particles, more than 70 mg/L lead to cell death after 24-h incubation, as the fullerene concentration and incubation time increases, the cell mortality increases (Han and Karim 2008). Numerous reports focused on in vitro and in vivo cellular uptake and toxicity evaluation of fullerene and its derivatives (Sayes et al. 2004, 2005; Bosi et al. 2004; Jia et al. 2005; Yamawaki and Iwai 2006; Porter et al. 2006; Han and Karim 2008; Kim et al. 2010; Cha et al. 2012). Jia et al. reported that unmodified C60 showed lower cytotoxicity than single- and multi-wall carbon nanotube to macrophages (Jia et al. 2005). The cellular uptake of human macrophages by C60 was very low, as low as that of treated SWCNTs, and their cytotoxicity was lower than that of graphite (Fiorito et al. 2006). Fullerene toxicity was evaluated in vivo on rats and fishes. Chem et al. demonstrated polyalkylsulfonated C60 dispersion orally, intraperitoneally, and intravenously, the results revealed no lethal damage was observed by oral administration and the median lethal dose (LD50) was approximated as 600 mg/kg in intraperitoneal administration. Nephropathy was induced after accumulation of materials in kidney by intraperitoneal and intravenous administration. The toxicological studies using fishes (Zhu et al. 2007; Usenko et al. 2008) showed C60 can act as a pro-oxidant and educe a toxic response via oxidative stress, considerable oxidative damage to brain lipids was reported and decrease in the hatching rate of zebra fish embryo and fin distortion. In summary, the toxicity parameters (shape, size, and oxygen species) of fullerene nanoparticles are still being evaluated. Major drawbacks are lipid peroxidation with functional groups mediating the toxicity and their accumulation in organs mainly in liver, due to massive binding with plasma protein and hindering their response for drug delivery applications. Furthermore, fullerene toxicity could be ascribed to their interaction with light (Prylutska et al. 2006), therefore, high controlled toxicity analysis of fullerene nanoparticles protected from light exposure, can evaluate the true toxic nature of fullerenes for medical applications. Carbon nanotubes in both forms, single- and multi-wall CNT, are widely considered to be as speculate materials for biomedical applications. The large length-to-diameter aspect ratio (more than 106; 1 mm length and 1 nm diameter) with high surface area makes it amenable for high molecular detection and recognition. Pristine CNTs are intrinsically insoluble in water and biological media, i.e., they cannot be explored directly in drug and bimolecular fields. Thus, CNTs should be functionalized to render them soluble and compatible in cell culture media. The toxicity of single- and multi-wall CNTs is attributed to numerous factors, for instance, length to diameter ratio, assay methods, functionalization, concentration, time of exposure, and nature of cells investigated. Wörle-Knirsch et al. found that the toxicity of SWCNTs depends on the preparation assay; the results showed that the viability of A549 decreased up to 40 % via MTT assay after addition of SWCNTs, but no changes on viability by using WST-1 assay method (Wörle-Knirsch et al. 2006).

Conclusion and perspectives

There is growing interest in GO as a material for biomedical application, in particular as a platform for drug delivery. This is due to the rich functional groups on the surface of the material which allow further facile functionalization to improve biocompatibility, dispersion, and of course drug loading. Within the review, the structure and properties of GO that allow covalent and non-covalent functionalization have been discussed. Toxicity investigations show functionalized GO to be biocompatible. Various works demonstrate GO derivatives are exciting nanocarriers for the loading and delivery of therapeutic drugs.