Surface functionalisation significantly changes the physical and electronic properties of carbon nano-dots

The electronic structure of glycan functionalised carbon nano-dots is greatly affected by the choice of carbohydrate.


General
Chemicals were purchased and used without further purification. Core CD formation was conducted in a domestic microwave (Tesco Homebrand) 700W. Concentration centrifugation tubes were GE Healthcare Life Sciences VIVASPIN 20 with a 10 kDa molecular-weight cut off (MWCO) filter. Kochetkov's amination of carbohydrates was conducted using Biotage Initiator+ microwave reactor. 0.5-01 kDa MWCO Biotech Cellulose Ester membranes were used for dialysis. Extracts were concentrated under reduced pressure using both a Büchi rotary evaporator at a pressure of either 15 mmHg (diaphragm pump) at room temperature. 1 H and 13 C HSQC NMR were measured in D2O at 500 MHz. All spectra were taken on a Bruker Advance III HD 500 Cryo. 1 H and 13 C NMR chemical shifts are quoted in parts per million (ppm) and referenced to the residual solvent peak (D2O: 1-H = 4.70 ppm). Coupling constants (J) given in Hertz. Multiplicities are abbreviated as: s (singlet), d (doublet), t (triplet), q (quartet), p (pentet) and m (multiplet). Assignments were made with the aid of HSQC NMR experiments. FTIR was conducted on a Bruker ATR. Zeta analysis was carried out using Malvern Instruments Nano-Z ZEN 2600. Fluorescence measurements were taken on a Perkin-Elmer LS45 in quartz cuvettes purchased from ThorLabs. Absorbance measurements were conducted on Cary UV-Vis 50 spectrophotometer in quartz cuvettes. All measurements were made in HPLC-grade methanol. Before any of analysis the CDs were sonicated for a minimum of 30 minutes. A single batch of CD cores was used for all experiments. All data analysis was conducted using Matlab unless otherwise stated.

Microwave assisted Kochetkov's amination
Anhydrous DMSO (0.8 mL), ammonium carbonate (1g) and sugar (250 mg) were mixed together in a microwave vial and sealed. The vial was then heated to 60˚C for 90 minutes in a reaction microwave (10 W, 250 psi). The resulting solid was dissolved in H2O and freeze dried.

Core synthesis
Glucosamine•HCl (1.00 g) and 4,7,10-trioxa-1,13-tridecanediamine (TTDDA) (1.35 ml) were mixed with 20 ml double-distilled water (ddH2O). This mixture was then heated in a domestic microwave (3 mins, 700 W). The resulting brown oil was washed and sonicated with CHCl3 several times, discarding the supernatant, until the supernatant is clear. The brown oil was then dissolved in 20 ml ddH2O and centrifuged through GE Healthcare Life Sciences VIVASPIN 20 with a 10,000 Da MWCO filter (5000 rpm, 1 hour). The sample was then passed through a 200 nm syringe filter and dialysed overnight against ddH2O at 0.5-01 kDa MWCO with Biotech Cellulose Ester membranes, replacing the ddH2O after the first couple of hours. The resulting solution was then freeze-dried to yield approximately ~1 g of core carbon dots.

Glycan functionalisation
The carbon dots were dissolved in methanol to a concentration of 10 mg.mL -1 and sonicated. This solution was then passed through a 200 nm syringe filter and mixed with 0.5 equivalence by weight of succinic anhydride. The solution was subsequently stirred vigorously overnight. The resulting solution was then reduced and the brown oil washed and sonicated with tetrahydrofuran several times, discarding the supernatant, until the supernatant was clear. The resulting brown oil was then dissolved in methanol, reduced in vacuo and weighed. The brown oil was then dissolved in ddH2O to a concentration of 10 mgmL -1 and sonicated for 5 minutes. This solution was then passed through a 200 nm syringe filter and mixed with 2 equivalents by weight of carbonyldiimidazole (CDI) and either 10 equivalents by weight of aminated monosaccharide or 20 equivalents by weight of aminated disaccharide to ensure homogeneous glycan functionalisation. This solution was then stirred vigorously overnight at room temperature. The solution was then freeze dried and the resulting brown oil was dissolved in ddH2O and dialysed overnight against ddH2O using Biotech Cellulose Ester membranes (at 0.5-01 kDa MWCO), replacing the ddH2O after the first couple of hours. The resulting brown oil was then again freeze-dried. For storage the carbon dots were dissolved in either HPLC-grade methanol and kept at 4 ˚C to prevent aggregation.
The different glycan functionalisations were performed on the same batch of acid coated CDs, with all glycan conjugations performed with an excess of 1aminoglycoside to ensure all the acid groups reacted. This ensured that any difference observed in the homogeneity of glycan corona after glycan conjugation is attributed to the nature of the different glycans.

Choice of Carbohydrates
Carbohydrates were chosen for their biological relevance and applications, as discussed in the main manuscript. A range of monosaccharides and disaccharides were used to explore the effect of different hydrophobic and hydrophilic regions formed by the different orientation of hydroxyl functional groups, although no trends emerged from our studies. Other carbohydrates such as trisaccharides (e.g. maltotriose) or polysaccharides (e.g. dextran) were not used to functionalised the CDs due to their aggregation in water and their low solubility in less polar solvents.
Carbohydrate-carbohydrate 1 and carbohydrate-protein 2 interactions have been suggested as mediators of important biological processes such as cell adhesion and aggregation, etc. These interactions are rather weak in a monovalent setting, but become significant when in a multivalent environment. Fundamental studies in the field to try to understand the nature of these key interactions have shown that carbohydrates can interact via hydrogen bonding through the OH groups and the apolar face in the sugar scaffold can also facilitate CH-π stacking interactions with aromatic motifs. 3,4 These interactions are dependant on the type of glycan and the glycan 3D surface presentation. It could be proposed that different glycans can thus interact with the surface of the CD they are conjugated on, which bears sp 2 motifs, 5 in a different manner depending on the nature of the sugar. Also, interactions between each class of glycan within each CD will be different depending on the sugar and therefore their 3D presentation would be affected.

CD stability
The core-CDs were also investigated by silica chromatography and size-exclusion chromatography (G10 Sephadex, Sigma). In both cases fluorescence and nanoscale structure could not be separated (observed by fluorescence spectroscopy and TEM). Previous work has also shown that the CDs are stable in a range of pH from 0-13 and when exposed UV light for a prelonged period. 5

NMR spectroscopy
The 1 H and HSQC 500MHz NMR spectra of the CDs are shown below in figures S1-S6. All samples were dissolved at 5 mg/mL in 0.8 mL of D2O using Norrell Select Series 7" NMR tubes (S-5-500-7). All spectra were taken on a Bruker Advance III HD 500 Cryo. All shifts are quoted in ppm. Residual internal D2O is at δ= 4.70. Peaks were identified using both 1 H and HSQC NMR.
The 13 C and 1 H NMR spectra of the core CDs shown in Figure S1, and reproduce our previously reported data. 5 The peaks arising from the TTDDA linker are identified and assigned. Additional peaks to the TTDDA are also observed at the following ranges: hydrogen-1 δ= 7-9 and C-13 δ= 115-150. These peaks are produced by hydrogens on aromatics and phenols on the surface of the CDs. C=C bending and stretching, Aryl C-O stretching and a weak phenol O-H bending signal can also be observed in the FTIR to support this conclusion.
The NMR spectra for the glycan-functionalised CDs are shown in figures S2-S6. For each of these spectra the TTDDA linker is identified, although often not completely assigned due to overlapping peaks from the functionalised carbohydrates (hydrogens attached to carbons 2-6). These peaks are themselves also difficult to assign individually due to their overlap, however appear in the following characteristic ranges: 1 H δ= 3.0-3.8 and C-13 δ= 55-80. Due to the apparent congestion in the NMR spectra, the successful carbohydrate functionalisation was instead determined by the identification of the characteristic doublet peak associated with the hydrogen on the anomeric carbon of intact carbohydrates and which is shifted when compared to the unconjugated glycoside. This doublet is found in the following ranges: 1 H δ= 4.2-5.5 and 13 C δ= 90-105. Often more than one doublet is observed, which arises from reduced conformational freedom that the carbohydrate experiences due to being tethered to the nanoparticle surface. This was observed for all of the glycan-functionalised CDs, and agrees with our previous NMR studies of lactose-CDs and work on the glycanfunctionalisation of CdSe quantum dots. 5,6 For each of the CD species, peak assignments are given in the figure caption and labelled with reference to the respective skeletal CD-glycan structures.
To further confirm successful glycan functionalisation of the nanoparticles, Diffusion-Ordered NMR Spectroscopy (DOSY), which probes the diffusion coefficient for each of the components of the 1 H NMR spectrum, was acquired. The anomeric peaks of each of the sugars were observed to have similar diffusion coefficients to the chemical shifts associated with the CD core, confirming that the glycosides are attached to the CDs. The DOSY data with their accompanying 1 H NMR spectra are shown in figures S7-S12. The DOSY data was analysed using DOSY Toolbox produced by Nilsson et al.. 7 The effective hydrodynamic radius of the CDs is also calculated from the diffusion coefficient using the Stokes-Einstein equation: Where D is the diffusion coefficient, kB is the Boltzmann constant, T is the temperature, η is the dynamic viscosity and r is the effective hydrodynamic radius of the particle. This should not be considered as an accurate measure of particle diameter as the particle is assumed to be spherical.
All of the DOSY peaks corresponding to the glycan functionalisation on the CDs were observed to have smaller diffusion coefficients than free sugars in solution therefore are attached to a particle of larger volume. A table of literature values of the diffusion coefficients for carbohydrates as well as the values for the DOSY peaks associated with the glycan-CDs are given in Table S1 for comparison.  The ratios of the components of the glycan-CDs was also investigated by NMR, this is shown in Figures S13-S14 and Table S2.

Fourier transformed infrared spectroscopy (FTIR)
The CDs were investigated using FTIR. The spectra support the earlier findings of NMR spectroscopy and shown in Figures S13-S18.

Glycan-functionalisation quantification by the phenolsulfuric acid assay
The glycan functionalisation was quantified using the phenol-sulphuric acid assay previously developed by others. 13 The calibration curve for the different carbohydrates is shown in Figure S19.

Figure S19: The absorbance calibration plots for the assay. The concentration of carbohydrate refers to the concentration at which it was added to the mix. The peak absorbance was identified at 490nm for all the carbohydrates.
The concentration of glycan was then determined using the same assay. This is given in Table S2. There are several additional sources of error: (i) the observed glycanconcentration is dependent on the ease of cleavage of the glycan from the CDs, which will be affected by the presentation of the glycan on the CD, which our AFM measurements demonstrate is inhomogeneous as discussed in the main manuscript; (ii) and a signal was also observed for the core-CDs however it is unclear whether this signal will remain for the glycan-CDs.

Transmission Electron Microscopy (TEM)
TEM was performed on samples drop cast onto 4 nm thick carbon-coated 3mm-copper grids. The samples were prepared in HPLC grade H2O at 5 mg/ml except for the lactose and core CDs which were prepared at 1 mg/ml and were dropped five times onto the same grid. The samples were all imaged on a Jeol 2100 TEM instrument at a voltage of 200kV. We were unable to attain sufficient contrast to directly image the glycanfunctionalisation or the aromatic domains of the CDs via TEM measurements, due to the low electron density of surface carbohydrates as opposed to crystalline carbon. This is visualised in the imaging of the core as the crystalline lattice appears to span the particle, suggesting only the crystalline domain provides sufficient contrast ( Figure  S26). For functionalised CDs a slight increase in the radius, is however, still observed. The size distributions of each of the CDs are given in figures S20-S25. The size distributions agree with previous TEM measurements. 5 It should be noted that TEM is intrinsically an underestimate of particle size due to the low electron scattering cross section of the non-crystalline carbon regions, with previous studies reporting aromatic surface coverage. 5 For the core-CDs it is possible to image the crystalline lattice, ( Figure S26). The lattice spacings observed by line profiles (Table S3) correspond to both sp 2 (100) and (020) or sp 3 (111) and (110), however for sp 2 carbon a lattice spacing of 0.33 nm corresponding to the (002) would also be expected. Such a conclusion is in agreement with our previous observations. 5 Example images of each of the glycan-CDs are shown in figures S27-S31.    5 It should be noted that the crystalline lattice spans the width of the particle, giving evidence that only the crystalline carbon is imaged with TEM.

Small angle X-ray scattering (SAXS)
Small-Angle X-Ray scattering was conducted using a SAXSLAB Ganesha 300XL with a Pilatus 300K detector and an evacuated sample chamber. All samples were prepared in HPLC grade methanol and measured in 1.5mm borosilicate glass capillaries (Capillary Tube Supplies) sealed with UV-curable adhesive (Norland Optical Adhesive). Samples were run over a q range of 0.015-0.65 Å -1 . Mannose, lactose and maltose CDs had an exposure time of 60 minutes; the core, glucose and galactose samples had an exposure time of 30 minutes. Samples were background corrected with HPLC grade methanol. Initial data processing was carried out using SAXSGUI; Guiner analysis was undertaken using Matlab. From previous TEM studies, 5 the CD particles are known to have approximately 2 nm core; this was used to define the Guinier region.
Guinier plots for each CD species are given in Figure S32. The raw data is shown in Figures S33.

Guinier analysis
The Guinier approximation was used to calculate the radius of the CD particles, and is found by the following approximation: (2) Where I is the intensity, Rg is the radius of gyration and q is the scattering vector is defined as: Where is the scattering angle is the wavelength of the source. The Guinier region is defined as when (with q in Å -1 and Rg in Å): Remembering that for a sphere: Where R is the radius. So for a sphere of predicted radius of ~2 nm, the range the Guinier approximation applies to the range: The background methanol was subtracted and the data was used for the range 0.001 Å -2 < q 2 < 0.007 Å -2 , which returned the following results in figures S32. " #### is used to give the "adjusted r-squared" statistic.

Atomic Force Microscopy (AFM)
AFM measurements were collected utilising a Multimode VIII microscope with NanoscopeV controller operating with PeakForce feedback (Bruker, CA, USA). SCANASY ST-FLUID+ cantilevers were used with a nominal tip radius < 2 nm and spring constant of 0.7 Nm -1 . The samples were prepared by drop casting 15 µl of 0.1 pgml -1 CDs in HPLC-grade methanol onto freshly cleaved mica; the excess methanol was blown off with nitrogen. The samples were allowed to dry for a minimum of 20 minutes before AFM experiments under ambient conditions.
For each CD sample, a single CD was imaged at high resolution to provide a detailed image of topography and adhesion. The tip-surface adhesion imaging gives further significant insights into the glycan-CD structure. For each glycanfunctionalisation, several additional CDs have been imaged at lower resolution.
Tip-surface adhesion imaging gives further significant insights into the glycan-CD structure. These adhesion data reveal that the core is a low adhesion centre (dark blue), consistent with crystalline or aromatic carbon. The AFM measurements also show regions of high adhesion areas surrounding the core (light blue) associated with the glycan-corona, as expected for dense regions of the functionalised carbohydrates.
For the core CDs, we collected a large amount of data to investigate the typical core size distribution using AFM. The results are displayed as a histogram in Figure  S34. Images of the core and glycan-functionalised CDs are given in figures S35-S39; the increase in adhesion for the glycan-CDs compared to the background is given in figure S40; a topographic map of the tip used to calculate the effective tip radius is given in Figure S42 and sample topographic profiles of each of the CDs are given in Figure S43.  Figure S32. This data suggests a larger radius than previous TEM measurements. This is because TEM is most sensitive to crystalline regions, and may provide poor contrast for the surface functionalisations. Further, and for reasons given in the main manuscript AFM intrinsically over-estimates the size-distributions of particles sized <~ 2nm. It is also likely that the AFM includes some small aggregates where the sub-structure could not be resolved.

Absorption Spectroscopy
The ultraviolet-visible absorption spectra of CDs were recorded using a Cary UV-Visible 50 spectrophotometer. The absorption spectra are dominated by bands at 273 and ~200 nm which are invariant to CD functionalisation. The absorption spectra at wavelengths longer than 315 nm are far more sensitive to glycan conjugation. This variation with functionalisation, which we also observe in the 2D fluorescence spectra, is ascribed to a modification of the electronic states which must be physically located on the surface of CDs. These most probably involve the aromatic shell that surrounds the crystalline sp 3 carbon core of the CDs.
TA experiments employed 340 nm pump excitation, and therefore excite a both core and surface states which have overlapping absorptions in this region. Figure S44: UV-visible absorption spectra for the CDs in methanol between 200 and 800 nm. All of the spectra were taken at a concentration of 1.0 mg/ml in methanol in a 3mm path length quartz cell.

Fluorescence Spectroscopy
Below are the fluorescence spectra for the CDs. All fluorescence measurements were recorded using a Perkin-Elmer LS45 spectrometer. All spectra were acquired for 1.0 mg/ml CD in methanol concentrations in a 3mm path length quartz cell.
The 2D fluorescence excitation-emission correlation spectra display at least one non-Kasha type fluorescence pathway at lemission ~430 nm, which is blue-shifted relative to the main fluorescence band at ~ 465 nm. The intensity of the blue-shifted feature is minimal for core CDs. The intensity of this feature is far more prominent for functionalised CDs, but varies with the precise carbohydrate. This feature is correlated with the changes observed in the linear absorption spectra at l > 315 nm, and from its position in the 2D fluorescence spectrum means excitation and emission are disconnected from the other absorption bands. Figure S45: Two-dimensional excitation-emission fluorescence spectra of CD samples.
The quantum yield of the CDs was calculated using quinine sulphate as a reference. The data from the analysis is given in Figure S46 and the quantum yields tabulated in Table S4. Figure S46: Fluorescence quantum yield measurements for the CDs. The fluorescence was integrated across the emission spectrum at a 340nm excitation. All measurements were made in 0.1M H 2 SO 4. Quinine sulphate was used as a control (known quantum yield =0.54) quantum yields are given in Table S4.

Transient Absorption Spectroscopy
A Ti:Sapphire oscillator (Vitara-S, Coherent) was used to seed a regenerative amplifier (Legend Elite HE+, Coherent, ~40 fs, 800 nm) which was used to pump an optical parametric amplifier (OperA Solo, Coherent) generating 340 nm pump pulses (~100 fs). White light supercontinuum probe pulses (350-630 nm) were generated by focusing a very small portion of 800 nm light into a rastered CaF2 plate (3 mm thick). At the sample position, the pump fluence was 600 nJ in a focal spot size of ~250 µm. The white light probe was focussed and overlapped with the 340 nm pump at the sample, with the transient absorption signals emitted collinearly to the probe beam. The probe and signal beams were collimated and imaged into a spectrograph (Shamrock 163, Andor) coupled to a 1024 element CCD array (Entwicklungsbüro Stresing). A copper sulfate solution was used to remove residual 800 nm from the white light supercontinuum before imaging into the spectrograph. The pump-probe time delay, t, was controlled with a mechanical delay stage (DDS220/M, Thorlabs). For all measurements the relative orientation of pump and probe pulses was set to magic angle. Data acquisition was performed in Labview. Transient absorption data were postprocessed to compensate for the chirp in the white light probe using the Matlab suite. The global analysis was performed in Glotaran.
The concentration of carbon dots used in TA experiments was 1 mgmL -1 , corresponding to an absorbance of ~0.3 at 340 nm. CD samples were continuously flowed throughout TA experiments in a 1 mm path length Starna flow cell. Figure S47: Chirp-corrected TA spectra for (a) core, (b) galactose, (c) lactose, (d) maltose, (e) glucose (f) mannose samples.

Global Analysis
A global analysis of the transient absorption data was performed in the Glotaran software package. 14 Such analyses avoid the problems associated with overlapping transient features and fits to data acquired at single wavelengths. The most reliable and physically plausible fits were returned by a sequential kinetic scheme (i.e. A ® B ® C) as opposed to a scheme that involved parallel decay pathways. These analyses decompose the transient absorption spectra into evolution assisted decay spectra (EADS) with associated concentration profiles. Unlike in a target analysis, the EADS merely reflect which spectral components decay that decay at the same rate, with an associated concentration profile. Each TA dataset for the various CD samples were modelled with three EADS components and returned very similar spectral profiles.
Examples of fits to data in the frequency and time domains are given below for core CDs.