Chiral nanoprobes for targeting and long-term imaging of the Golgi apparatus† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c7sc01316g Click here for additional data file.

The targeting and long-term imaging of the Golgi apparatus have been realized via l-cysteine functionalized nanoprobes.


Introduction
Subcellular targeting strategies have redened the frontier of life processes as well as drug design. 1,2 As a eukaryotic organelle, the Golgi apparatus is essential for biogenesis, secretion, and intracellular distribution of a wide range of macromolecules. 3 It has been reported that morphological changes of the Golgi are related to external stimuli, 4 thus can effectively reect the physiological state of cells. With the development of the membrane fusion method using N-[7-(4-nitrobenzo-2-oxa-1,3diazole)]-6-aminocaproyl sphingosine (NBD C 6 -ceramide) for exploring the sphingolipid transport and metabolism mechanism, 5 ceramide analogues have become one of the most widely used commercial dyes for labeling the Golgi. 6 Due to the lack of specicity of ceramide analogues, however, the plasma membranes as well as the mitochondria are also labeled eventually besides the Golgi apparatus. Other Golgi stain approaches, 7,8 using antibodies, for example, also lack longterm imaging ability, making it difficult to observe real-time changes of the Golgi at the single-cell level under continuous laser excitation. Therefore, it is highly desirable to design and develop new optical probes for the Golgi with a prolonged imaging ability.
Fluorescent carbon quantum dots (CQDs), since their rst discovery in 2004, 9 have been found to be physicochemically and photochemically stable without photobleaching, making them a star family in the biosensing and bioimaging elds. [10][11][12][13] However, bare CQDs are usually weakly uorescent and it is hard to attach functional groups on the surface. 14,15 Regardless of continuous efforts to circumvent the problems, 16 challenges still remain to engineer CQDs with desirable biosensing and labeling properties, such as a high quantum yield and preferable selectivity for the long-term and real-time monitoring of targets of interest. 17 Herein, we report a novel type of CQD for the long-term in situ imaging of the Golgi apparatus. It has been reported that galactosyltransferase and protein kinase D are capable of anchoring in the Golgi apparatus via their cysteine residues or cysteine rich domain, 18,19 which inspires us to combine the principle of the Golgi localization of proteins and carbon nanotechnology to develop an optical probe for Golgi targeting and imaging. We synthesize novel uorescent CQDs with abundant cysteine residues and an L-type spatial structure using a pyrolysis method with citric acid and L-cysteine as the carbon sources and controlling the pyrolysis temperature. The as-prepared LC-CQDs exhibit excellent long-term Golgi targeting and imaging capabilities that could be attributed to their high quantum yield (68%) and photostability as well as their good biocompatibility. This dependence of the targeting of the Golgi on L-cysteine is further proven by using cysteine modied uorophores and silica nanoparticles. This study provides an effective method for Golgi targeting and imaging over a long time scale, which may be applied in drug-delivery and therapy, as well as in evaluating disease progression occurring within the Golgi.

Results and discussion
Characterization of L-cysteine-rich CQDs (LC-CQDs) The as-prepared LC-CQDs eventually are 8.5 AE 3.5 nm in diameter and are highly uorescent with an absolute quantum yield of 68% ( Fig. 1a and b and S1-S4 †). The free thiol group of L-cysteine is well preserved on the surface of the LC-CQDs, as indicated by the vibrational frequency (-SH, 2565 cm À1 ) of the free thiol groups ( Fig. 1c and S5-S10). 20 The average number of L-cysteine residues on each LC-CQD is 248 (calculation details are in the ESI, Scheme S1 †). Interestingly, the as-prepared LC-CQDs exhibit strong circular dichroism signals at 245 nm and 350 nm (Fig. 1d), which are signicantly different from those of the L-cysteine precursor (Fig. S11 †). The developed chiral centers, probably arising from chiral imprint or chiral induction, 21,22 are preserved in the carbonization process. As is evident from Fig. 1d, LC-CQDs and D-cysteine-rich chiral CQDs (DC-CQDs) display opposing Cotton effects in the range of 200-400 nm, reinforcing the existence of the chirality in both LC-CQDs and DC-CQDs. Thus, it can be inferred that the chirality of the as-prepared LC-CQDs and DC-CQDs is transferred from cysteine to the CQDs.
LC-CQDs display excellent uorescence properties. As Fig. 1e shows, the normalized UV-FL spectra of the LC-CQDs possess clearly resolved absorption peaks and symmetrical FL peaks, and the blue emission of the LC-CQDs has a maximum wavelength (l max ) of 420 nm, giving an absolute quantum yield of 68%. The strong uorescence emission of the LC-CQDs is attributable to the co-doping effect of N and S. 23 Signicantly, the LC-CQDs exhibit superior photostability compared to the FITC dye, uorescein and the CdTe QDs (regarded as photostable uorescent labels), as Fig. 1f shows. The uorescence of FITC, uorescein and the CdTe QDs was found to be quickly quenched in 30 min under UV irradiation due to severe photobleaching, but of the uorescence of the LC-CQDs, about 95% of the initial intensity was preserved aer 5 h of UV irradiation ( Fig. S12 †), thus the LC-CQDs were extremely stable in contrast to FITC, uorescein and the CdTe QDs.

The Golgi targeting ability of the LC-CQDs
The specic Golgi targeting ability of the as-prepared LC-CQDs is demonstrated by co-staining human epithelial cells (HEp-2 cells) with N-acetylgalactosaminyltransferase-GFP (Golgi-GFP) and Bodipy ceramide, the two frequently adopted probes for Golgi-staining. Obvious uorescence in the Golgi is observed (Fig. 2a) aer 4 h of incubation of HEp-2 cells with the LC-CQDs (0.5 mg mL À1 ). The uorescent area matches very well with those of Golgi-GFP and Bodipy ceramide ( Fig. 2a-d) with a Pearson's correlation factor higher than 0.9, indicating that preferential accumulation of the LC-CQDs in the Golgi has occurred ( Fig. S13 and S14 †). Similar results were also obtained for A549, HepG2, and Hela cells (Fig. S15 †). The intracellular uptake and transportation results indicate that the uptake of LC-CQDs is an energy-dependent process mediated by clathrin and caveolae; large quantities of LC-CQDs are transported via early endosomes and late endosomes and are eventually transported to the Golgi through the retrograde trafficking route (Fig. S16-S19 †). Immunouorescence results conrm that the LC-CQDs can stain both the cis-Golgi and trans-Golgi ( Fig. 2e-l), and the transmission electron microscopy results show that the LC-CQDs are inside the lumen of the Golgi (Fig. 2m-p).

Long-term in situ Golgi imaging ability of the LC-CQDs
To examine the suitability of the LC-CQDs for the long-term in situ imaging of the Golgi, different properties, such as imaging duration, biocompatibility and photostability, were evaluated. The cell number increases from four to seventeen, due to cell proliferation ( Fig. 3a-e), when HEp-2 cells are incubated with the LC-CQDs for 4 days. In contrast, 1 mM NBD C6-ceramide, whose concentration is considerably lower than that used in typical applications, induced cell rupture and detachment aer  On the other hand, the uorescence of the LC-CQDs in the cells remained higher than 80% in the Golgi even under continuous laser excitation for 3600 s ( Fig. 3f and j), indicating that the LC-CQDs have good photostability, which is better than those of Bodipy ceramide and Golgi-GFP as the uorescences of Bodipy ceramide and Golgi-GFP disappear within 600 s (Fig. 3g, h, k and l). The sequential 3D imaging of HEp-2 cells using spinning disk confocal microscopy also demonstrates the superior photostability of the LC-CQDs. For Bodipy ceramide and Golgi-GFP labeling, 70% and 90% loss of uorescence aer 80 3D 'stacks' are observed, respectively ( Fig. S26b-d †). As a clear contrast, the uorescence signal of cells treated with the LC-CQDs decreases by only 10% aer the same number of 3D 'stacks' (Fig. S20a †). Furthermore, the results of the 2 0 ,7 0 -dichlorouorescin diacetate (DCF-DA) assay demonstrate that the LC-CQDs would not generate reactive oxygen species during long-term in situ Golgi imaging (Fig. S27 †). These observations illustrate that the LC-CQDs are feasible for the in situ long-term uorescence observation of the Golgi.

The Golgi targeting mechanism of the LC-CQDs
Both the L-type stereo conguration and the existence of free thiol groups are proven to be necessary for targeting the Golgi apparatus. Compared with LC-CQDs, DC-CQDs prepared from D-cysteine and citric acid ( Fig. 1d and S28-S30 †) exhibit a limited Golgi targeting ability (Fig. S31a †) with a Pearson's correlation factor of 0.35, indicating that the chirality of the CQDs strongly inuences the targeting ability towards the Golgi. CQDs prepared from precursors without chiral centers,  e.g. thioacetamide, are found to randomly spread in cells with a Pearson's correlation factor of 0.28 (Fig. S31e and S32-S34 †), reinforcing the signicance of a proper stereo structure of the CQDs for Golgi targeting. Furthermore, the existence of free thiol groups is proven to be essential for Golgi targeting. Carbon dots prepared from precursors carrying no free thiol groups, Smethyl-L-cysteine (Fig. S35-S37 †) or L-alanine (Fig. S38-S40 †), are found to have very low Golgi targeting abilities as indicated by Pearson's correlation factors of 0.21 and 0.17 ( Fig. S31c and  d †), respectively. A further competitive inhibition experiment shows that molecules without thiol groups cannot inhibit the targeting ability of the LC-CQDs or compete with the LC-CQDs in binding to the Golgi as L-cysteine does (Fig. S41 †). These observations solidify the necessity of free thiol groups for the Golgi targeting and also suggest that the LC-CQDs may bind to the sulydryl receptor site of the Golgi through the formation of disulphide bonds in the oxidizing environment of the Golgi lumen. 24

L-Cysteine induced Golgi targeting
In order to demonstrate the feasibility of L-cysteine in Golgitargeting, we link L-cysteine with different materials using chemical approaches. As Fig. 4a shows, HEp-2 cells are incubated with L-cysteine functionalized uorescein and a complete colocalization is observed for the uorescein-cysteine molecules and Bodipy ceramide with a Person's correlation factor greater than 0.94 (Fig. 4b-d and S42-S51 †). In contrast, uorescein without cysteine functionalization is found to readily spread over the whole cells (Fig. 4e-h). Similar results are observed for L-cysteine modied silica nanoparticles with sizes of 40 nm and 80 nm (Fig. 5), further strengthening the hypothesis that L-cysteine could target the Golgi. In order to identify the multivalent effect of L-cysteine, meso-tetra(4carboxyphenyl)porphine (TCPP) is linked with different numbers of L-cysteine residues. Unfunctionalized TCPP is found to spread over the whole cells, while single cysteine functionalized TCPP colocalizes well with Golgi-GFP ( Fig. 4i-m) aer 5 h incubation. Double L-cysteine functionalized TCPP needs only 3 h to match with Golgi-GFP ( Fig. 4m and n and S52-S58 †), suggesting that the Golgi targeting ability is closely associated with the number of L-cysteine residues and the multivalent effect of L-cysteine can improve the Golgi targeting ability. Similarly, the Golgi targeting ability of the LC-CQDs would improve with an increasing number of cysteine residues on their surface (Fig. S59 †). The multivalent effect of the LC-CQDs is also veried by the uorescence recovery aer photobleaching (FRAP) (Fig. S60 and S61 †). These results conrm that L-cysteine is an effective agent for specic Golgi targeting.

Viral infection induced Golgi fragmentation
The long-term in situ targeting ability of the LC-CQDs could be further demonstrated by visualizing the dynamic morphology changes of the Golgi at the single-cell level during viral infection. Stained with the LC-CQDs, the Golgi apparatus of HEp-2 cells emits blue uorescence (Fig. 6a). Aer the infection of HEp-2 cells with respiratory syncytial virus (RSV) for 12 h, the Golgi breaks into fragments and becomes scattered (Fig. 6b). The collapse becomes even more serious (Fig. 6c) aer 21 h. The Golgi morphology changes might be a more sensitive indicator of cell vital state changes, because the cell morphology does not show an obvious change during the Golgi fragmentation ( Fig. 6a-d). Similar morphological changes were also obtained for A549, HepG2, and Hela cells during the process of viral infection (Fig. S62 †). Identication of the viral attack-induced Golgi response reveals a transition stage from the living state to the death of the cells, thus providing a new and sensitive mode of investigating diagnostically relevant changes in the early stages of viral infection. To the best of our knowledge, this is the rst report about the Golgi response in the early stages of viral infection.

Conclusions
We have successfully demonstrated an approach to prepare chiral carbon quantum dots (L-type stereo structure, LC-CQDs) with abundant cysteine residues. The LC-CQDs exhibit unique properties, including a high quantum yield (68%), high photostability, as well as favorable biocompatibility. More importantly, the LC-CQDs exhibit a distinctive capacity for long-term in situ Golgi imaging via the L-type stereo structure and free thiol groups. Due to their long-term in situ Golgi imaging ability, the LC-CQDs are able to trace the real time process of intracellular injury, such as viral infection, by monitoring the morphological changes of the Golgi apparatus. Thus a transition stage from the living state to the death of the cells during the process of viral infection has been observed, providing an innovative mode of investigating diagnostically relevant changes in the early stages of viral infection. Furthermore, the exploration of cysteine for Golgi targeting provides an innovative subcellular targeting strategy for drug design and delivery, and thus illustrating that functional cellular CQDs that can target subcellular organelles such as Golgi body show high promise in the explorations of cellular activities of typical subcellular organelles.

Synthesis of the LC-CQDs
The LC-CQDs were prepared by directly pyrolysing L-cysteine and citric acid (CA). CA monohydrate (2 g, 9.5 mmol) was at rst placed in a 50 mL ask and heated at 200 C for 20 min, and then L-cysteine (1.5 g, 12.4 mmol) was added. Aer stirring with a glass rod, the yellow mixture was kept at 200 C for 60 min until a dark product was obtained. 2 mL of ultra-pure water was added to the mix of carbon quantum dots with cysteine and heating was continued if the yellow mixture did not turn dark. The dark product was nally neutralized with KOH solution (10 mL, 1.3 mol L À1 ). The LC-CQDs were collected by removing the large particles through ltering using 0.22 mm membranes, and then dialyzing against ultra-pure water for 48 h to remove small organic molecules and salts (3.5k MWCO, Spectra/Pro 3 dialysis sack). Other carbon quantum dots were synthesized by replacing the L-cysteine with D-cysteine, L-alanine, S-methyl-L-cysteine, thioacetamide and a gold-cysteine complex (1.5 g cysteine, 5 mL 5% HAuCl 4 ).

Synthesis of L-cysteine functionalized silica nanoparticles
Methanol (10 mL), ultra-pure water (3 mL), and concentrated ammonia (0.7 mL) were mixed at 30 C under stirring. Tetraethyl orthosilicate (200 mL) was then added to the solvent mixture followed by the addition of a uorescent complex (20 mL methanol, 0.16 mg uorescein isothiocyanate (FITC), and 4.8 mL 3aminopropyltriethoxysilane (APTES). The uorescent complex solution should be mixed and stored at 4 C for one day before  the synthesis process). Aer 4 h, the products were collected by centrifugation at 12k rpm and washed with ethanol. By adjusting the amount of concentrated ammonia used under the same conditions, the particle sizes of the silica nanoparticles were tuned to 40 nm (concentrated ammonia, 1.2 mL), and 80 nm (concentrated ammonia, 0.7 mL). Subsequently, the as-prepared silica nanoparticles were dispersed in 15 mL of ethanol followed by the addition of 200 mL APTES. Aer 12 h, the aminefunctionalized silica nanoparticles were collected and washed with ethanol. The amine-functionalized silica nanoparticles were then redispersed in 15 mL acetone followed by the addition of N,N 0 -disuccinimidyl carbonate (DSC, 300 mg) and N,N-diisopropylethylamine (DIEA, 200 mL). Aer another 12 h, the DSCfunctionalized silica nanoparticles were collected and washed with acetone. Finally, the DSC-functionalized silica nanoparticles were redispersed in 8 mL L-cysteine (Sigma-Aldrich) complex solution (8 mL ultra-pure water, 300 mg L-cysteine and 400 mL DIEA) for 4 h. The L-cysteine functionalized uorescent silica nanoparticles (LC-F-SiO 2 ) were collected by centrifugation at 12k rpm and washed with ultra-pure water. Some large products could be removed by centrifugation at 4k rpm.
Bodipy TR C 5 -ceramide (Thermo Fisher Scientic) or NBD C 6 -ceramide (Thermo Fisher Scientic) can be administered to cells as a complex with bovine serum albumin (BSA). The cells were washed until they were free of culture media and incubated with Bodipy TR C 5 -ceramide (5 mM) for 30 min at 4 C. The cells were then washed, incubated in 2% BSA solution for 90 min at 25 C, and nally observed under a uorescence microscope. CellLight® Golgi-GFP (Thermo Fisher Scientic) was added to the cells, incubated overnight, and the cells were ready to image in the morning.
The HEp-2 cells used in this experiment are the laryngeal cancer cell lines. The respiratory syncytial virus (RSV) used in this experiment is effective at infecting HEp-2 cells. In order to monitor the morphology changes of the Golgi during the viral infection, we selected Hep-2 cells as the host cells.
The data in the main text and ESI † were obtained from replicate experiments (n ¼ 5).

Electronic microscopy study of the LC-CQDs in the Golgi
Cells grown in cell culture asks were incubated with 1 mg mL À1 Au-doped LC-CQDs for 48 h, washed, and xed in 1% glutaraldehyde. The cells were then washed with 0.1 M PBS buffer and treated with 0.5% OsO 4 for 1 h at 4 C. Aer rinsing with ultra-pure water, the cells were stained with saturated uranyl acetate in 70% ethanol for 24 h at 25 C. The cells were then dehydrated in ethanol and embedded in Epon. Thin sections were obtained with an ultramicrotome. Electron micrographs were obtained with an electron microscope (Transmission Electron Microscope Hitachi-7500). Control samples were treated as described above, except that they were not incubated with LC-CQDs.

Immunouorescence study of the LC-CQDs in the Golgi
The cells were xed in 4% paraformaldehyde for 20 min at 25 C and permeabilized with 0.1% Triton X 100 for 2 min at 25 C. The primary antibodies were diluted 1/200 and incubated with samples for 1 h at 37 C. The secondary antibody was the goat polyclonal secondary antibody to Rabbit IgG-H&L (Cy3®) used at a 1/500 dilution for 1 h. 2% BSA was used for the blocking steps. All antibodies were purchased from Abcam.

Viral infection
Virus samples were diluted with RPMI 1640 and then introduced to the cells. The Golgi apparatus of the cells had been labelled by LC-CQDs before the addition of the virus. Aer incubating with RSV for 2 h at 37 C, the infected cells were cultured in RPMI 1640 supplemented with LC-CQDs and 2% FBS in a 5% CO 2 incubator. The morphology of the Golgi apparatus of the same cell was observed at different times.
HEp-2 cells used in this experiment were the laryngeal cancer cell lines. The respiratory syncytial virus (RSV) used in this experiment was effective at infecting HEp-2 cells.

Quantum yield
The absolute photoluminescence quantum yield was measured using Quantaurus-QY (Hamamatsu, Japan).