Carbon dot-based fluorescent antibody nanoprobes as brain tumour glioblastoma diagnostics

The development of efficient and sensitive tools for the detection of brain cancer in patients is of the utmost importance particularly because many of these tumours go undiagnosed until the disease has advanced and when treatment is less effective. Current strategies employ antibodies (Abs) to detect Glial Fibrillary Acid Protein (GFAP) in tissue samples, since GFAP is unique to the brain and not present in normal peripheral blood, and it relies on fluorescent reporters. Herein we describe a low cost, practical and general method for the labelling of proteins and antibodies with fluorescent carbon dots (CD) to generate diagnostic probes that are robust, photostable and applicable to the clinical setting. The two-step protocol relies on the conjugation of a dibenzocyclooctyne (DBCO)-functionalised CD with azide functionalised proteins by combining amide conjugation and strain promoted alkyne–azide cycloaddition (SPAAC) ligation chemistry. The new class of Ab-CD conjugates developed using this strategy was successfully used for the immunohistochemical staining of human brain tissues of patients with glioblastoma (GBM) validating the approach. Overall, these novel fluorescent probes offer a promising and versatile strategy in terms of costs, photostability and applicability which can be extended to other Abs and protein systems.


Introduction
Overall less than 20% of brain tumour patients are alive 5 years aer diagnosis, in part because they present late with large inoperable tumours. 1 There is an urgent need to develop new sensitive tests of brain tumours to help general practitioners in primary care. 2 The most common malignant primary brain tumour called glioblastoma is characterised by abnormal blood vessels resulting in a leaky Blood-Brain Barrier (BBB). 3 Glial Fibrillary Acid Protein (GFAP) is unique to the brain and not present in normal peripheral blood. Antibodies targeting GFAP are used to diagnose gliomas in tissue samples. There is evidence that GFAP crosses the leaky BBB and is an early non-specic peripheral blood biomarker which predates the clinical diagnosis of glioblastoma. 4 However, GFAP levels are too low for routine detection by commonly used protein diagnostic tests such as ELISA, and more sensitive methods for its identi-cation are needed. 5 Fluorescent labelling of proteins is a common strategy to investigate their role and function in cells, tissues and organisms. 6 Traditional efforts rely on the use of molecular dyes, which are usually expensive and predisposed to photobleaching. Alternatively, uorescent nanoparticles can be tuned to exhibit high stability, sensitivity and specicity for their desired target without the limitations of organic uorophores and as a result these nanoprobes have found many applications as more robust tools in the areas of bioimaging, drug delivery and diagnostics. 7 Among the different types of luminescent nanomaterials, carbon dots have emerged as a new class of carbon-based uorescent nanomaterials with semi-spherical morphology, unique optical and physico-chemical properties such as chemical inertness, high water solubility, resistance to photobleaching, low cost of fabrication, and very low cytotoxicity. 8 These carbon based uorescent nanomaterials have been hailed as alternative probes to semiconductor quantum dots which have been linked to heavy metal toxicity which restricted their use in vivo applications. 9 As a results, the use of carbonbased nanomaterials in biology as a platform for gene delivery, 10 cell imaging, 11 diagnosis, 12 and as theranostics 13 has raised a lot of interest.
CDs can be easily produced via the thermal degradation of readily available substances such as citric acid and ethylenediamine furnishing CDs with high uorescent quantum yield.
The uorescent excitation and emission pattern of CDs can be tuned by changing the synthetic conditions, variables such as the kind of solvent, temperature and the ratio of precursors used during the preparation may provide different nanoparticles able to emit in different spectral regions. 8b Therefore, the combination of these features makes of CDs excellent candidates for their use as uorophores for Abs labelling. However, despite the great advantages offered by CDs for many biosensing applications including examples for GFAP detection, 14 their use as uorophores for Abs labelling is still underdeveloped and Ab conjugation strategies to CDs have not been fully developed for direct clinical applications. 15 To this extent, we envisioned the use of uorescent carbon dots (CDs) for the development of cheap and photostable probes for Abs functionalization that can be used for the detection of GFAP in clinical samples.
Herein, we describe the development of a practical, low cost and general strategy for the labelling of Abs with uorescent CDs. This rst generation of Ab-CD conjugates combines EDC and strain promoted alkyne-azide cycloaddition (SPAAC) ligation chemistry to generate a new class of Ab-CD conjugates which are robust and photostable (Fig. 1). Moreover, the clinical versatility of the novel Ab-probes is demonstrated in the immunohistochemical staining of human brain tissues of patients with glioblastoma GBM.

Results and discussion
2.1 Synthesis of dibenzocyclooctyne (DBCO)-functionalised uorescent carbon dots The synthesis of dibenzocyclooctyne (DBCO)-functionalised CD 2 ready to be conjugated to azido-decorated Abs started from acid functionalised CD 3, which were prepared in one pot from citric acid and ethylenediamine under microwave irradiation (domestic microwave oven, 300 W) following a modied procedure by Mondal et al. 16 (Fig. 2A). The reaction mixture was dissolved in distilled H 2 O and precipitated in an excess of acetone several times to give acid-functionalized CDs, which  aer dialysis and centrifugal ltration (10 kDa molecular weight cut-off membrane) afforded blue emitting nanoparticles with a narrow size distribution as evidenced by uorescence measurements (Fig. 2B). Indeed, TEM revealed the presence of quasi-spherical nanoparticles with an average size between 2-5 nm (N ¼ 262) and a lattice interspacing of 0.34 nm (Fig. 2C) which correlate to a graphite core structure. 17 Functionalization of CD 3 with DBCO-linker 9, which was prepared in 4 steps and 47% overall yield, afforded DBCO-CDs 2 (Scheme 1). In brief, mono-amine protection of commercially available 4,7,10-trioxa-1,13-tridecanediamine 4 with Boc 2 O in CH 2 Cl 2 gave 5 18 in 99% yield. The free amine in 5 was then reacted with 4-nitrophenyl chloroformate to form activated carbamate 6, which could then be treated with commercial DBCO-amine 7 to give 8 in 55% yield over the 2 steps. Boc deprotection in the presence of TFA/CH 2 Cl 2 afforded 9 ready for CD conjugation. HATU mediated nanoparticle functionalization of acid coated CD 3 with 9 was carried out using a 1 : 0.5 w/ w ratio of 3:9, which was found to be optimum to ensure the nanoparticles remained in solution despite the hydrophobic coating. Furthermore, 1 H, HSQC and Diffusion Ordered (DOSY) NMR spectroscopy analysis demonstrated the successful conjugation of DBCO moieties on the CD (see ESI Fig. S11-S13 †). Moreover, the quantum yield measurements for CD 3 and DBCO-coated 9 was found to be 0.46 and 0.34, respectively ( Fig. S20 in ESI †).

Protein conjugation strategy
There are several antibody-drug conjugation strategies available which include: amide formation, periodate oxidation of carbohydrates in the Fc region and subsequent functionalization of the novel aldehyde function produced within the glycosidic moiety, click reactions involving 1,3-dipolar cycloadditions, thiol maleimide conjugations and thiol-mediated alkylation among others. 19 Among those, the EDC/NHS-promoted amide formation by targeting primary amines found on lysine residues, appears to be one of the most common and direct approaches for nanoparticle loading. 14a,20 Control over the Ab/ nanoparticle loading ratio to avoid the over functionalization of the Ab, as well as the type and length of spacer molecule between the probe and Ab are essential to avoid loss of protein's pharmacokinetic properties. Moreover, excessive labelling may also interfere with the antigen-biding event. In this context, an Ab/drug ratio of around 1 : 4 is found to be optimal 21 and this parameter was considered in the preparation of our Ab-CD conjugates. NHS activated conjugation represents the golden standard as it allows for a fast and reliable Abs covalent functionalization, NHS-functionalised molecules react with solvent exposed primary amines such as lysine residues on the surface of the Abs. 22 Taking into account the above considerations, NHS-linker 1 (for its preparation see ESI Section 3.3 †) 23 featuring an azido motif that can selectively react with the DBCO-CD 2. Moreover, both 1 and 2 were designed with a triethylene glycol spacer to help with water solubility and to reduce steric hindrance between the Ab and the surface of the CD.
2.2.1 BSA-CD conjugation strategy. With both CD-DBCO 2 and NHS-linker 1 in hand, the feasibility of our labelling approach was initially evaluated on bovine serum albumin (BSA) protein, as an inexpensive model system ( Fig. 2A). In the rst step, the protein was functionalised with the azidocontaining linker 1, in brief a solution of BSA in PBS (100 mL, 36.1 mM) was reacted with excess amounts of 1 (Fig. 3A, different molar excess of 1 were tested, entries a-f: from 40 to 1333 eq.) under gentle shaking at room temperature for 4 h. Removal of the excess of 1 and washes via spin ltration over 30 kDa cut-off membrane gave 10a-f with different degrees of functionalization from approximately 6 to a maximum of 32 linker units per protein as determined by Matrix Assisted Laser Desorption/ Ionization (MALDI) mass spectrometry using the spotting procedure described by Signor et al. 24 (see Fig. 3A, and ESI: Fig. S21 and Table S1 †).
Once the protein is decorated with the azido functionalities, chemoselective Cu-free click conjugation with DBCO-CD 2 was attempted in PBS by mixing 10a (BSA with 6 N 3 -linker units) and 2 at room temperature for 16 h. A 5-fold excess of 2 in weight with respect to the protein was used to ensure all the available N 3 moieties were conjugated. Following spin ltration (30 kDa cut-off membrane) to remove the excess of 2, BSA-CD conjugate 11a was obtained as determined by MALDI (Fig. 3B). The conjugation of CDs to BSA caused a shi toward larger molecular weights and a broadening of the peak as expected from protein conjugation with a disperse nanoparticle system such as our CDs. 25 Gel electrophoresis was also used to further conrm the effective BSA-CDs conjugation, by allowing us to compare the MW of the different protein adducts (Fig. 3C). Whereas the addition of 6 low molecular weight (MW) N 3 -linkers on 10a did not show any signicant changes on the gel when compared to with native BSA (Fig. 2C, i vs. ii), a noticeable increase in MW was shown for the BSA-CD adduct 11a (Fig. 3C, i vs. iii), which further validates the MALDI data. Gel electrophoresis analysis of BSA-N 3 derivatives 10b-f was also possible (see ESI Fig. S22 †).
Moreover, to exclude the possibility of non-specic BSA adsorption on CD nanoparticles, unfunctionalized BSA, which lacks azido motifs, and DBCO-CDs were pre-mixed together and  run on the same well showing no MW changes with respect to BSA alone (Fig. 3C, i vs. iv) which demonstrated the chemo-specic labelling of the protein via SPAAC reaction.

Anti-GFAP Abs-CD conjugation.
Having demonstrated successful protein labelling with our strategy, CD conjugation on clinically relevant rabbit polyclonal anti-glial brillary acidic protein antibodies (anti-GFAP Abs) was next attempted. NHS-linker 1 conjugation to anti-GFAP Abs was performed as before (Fig. 3A). A solution of Abs in PBS (11.1 mM) was treated with an excess of 1 (Fig. 4A, entries a-d: 0.12-2.41 mmol) and le shaking at room temperature over 4 h. MALDI analysis of the products showed peak broadening and shis towards higher MWs for the azido functionalised anti GFAP Abs which could be used to estimate the average degree of substitution (Fig. 4B, and ESI: Fig. S24 and Table S2 †). In general, it was found that a maximum of 30 azido containing linkers could be conjugated to the Abs at the higher concentrations, while a degree of functionalization of 4 linker moieties was achieved when 100 molar equivalents of 1 was used for the conjugation, which is optimum to maintaining Ab function and good pharmacokinetic and toxicology prole. 21 The anti-GFAP Abs-N 3 derivatives 12a-d were then treated with an excess of DBCO-CD 2 as previously described for BSA and following spin ltration over 50 kDa cut-off membrane to remove the excess of unconjugated 2, and anti-GFAP Abs-CD conjugates 13a-d were generated. Gel electrophoresis was used to conrm CD labelling of the anti GFAP-Abs. As for the BSA model gel electrophoresis showed negligible differences in terms of MW for 12a-d when compared to native Abs (Fig. S25 † i vs. iii-vi). It is worth noting that although MALDI clearly shows MS differences between native and azido functionalised Abs, no changes on MW were detectable on the gels for neither anti-GFAP Abs or anti-GFAP Abs-N 3 since the molecular weight differences between the species is negligible at the level of MW resolution for gel electrophoresis. Indeed, the different Abs-CD probes showed signicant increase in MW for 13a-d, which correlated to their degree of azido functionalization, when compared to native anti-GFAP Abs (Fig. 4C). 26 To conrm that the CD labelling of the Abs is not due to non-specic interactions, uorescence images of a native Abs and Abs-N 3 that were treated with DBCO-CD 2 prior to purication showed uorescent labelling only for azido containing Abs as expected (Fig. 4D). Moreover, as previously demonstrated for BSA, gel electrophoresis of unfunctionalized Abs were pre-mixed with 2 and run on the same well showed no MW changes with respect to Abs alone (Fig. S25, † i vs. ii) conrming the absence of non-specic interactions between the CD and the Abs.
Furthermore, western blot analysis using human GFAP with 13a (functionalised with 4 linker units) was used to demonstrate the novel anti-GFAP Ab-CD 13a adducts retained their ability to recognise the target antigen. To that end, a goat anti-rabbit secondary antibody equipped with a near IR probe was used on the western blot, conrming the presence of the rabbit anti-GFAP Abs bound to the human GFAP antigen (Fig. 4E). 27 2.3 Anti-GFAP Abs-CD immunostaining of clinical tissue brain cancer patient samples GFAP immunostaining is the most commonly used method to examine the distribution of astrocytes and the hypertrophy of astrocytes in response to neural degeneration or injury as in the development of glioblastoma. 4 To demonstrate the versatility of our CD-based Ab labels for diagnosis applications, we have examined GFAP in 13 formalin-xed paraffin embedded biopsy brain tumour samples from different patients (12 glioblastoma, IDH wildtype, WHO Grade 4 and 1 negative control schwannoma, WHO Grade I, see ESI: Table S3 †) using our conjugated antibody 13a (Fig. 5). We identied immunouorescence within all the glioblastoma cases (as assessed by a consultant neuropathologist KMK) using the conjugated anti-GFAP antibody 13a ( Fig. 5A and B, for the complete set of pictures see ESI Fig. S27 †). We identied the correct pattern of cytoplasmic staining (blue) of the GFAP intermediate lament in the glioblastoma cell cytoplasm (Fig. 5A and B). The intensity and extent of GFAP immunopositivity showed inter and intra-tumoural heterogeneity in keeping with known biological variation between cases. The negative control schwannoma showed no positive staining using the conjugated GFAP antibody as expected (Fig. 5C). There was no variation of GFAP staining with age, sex or molecular parameters within the small cohort as expected. In addition, control labelling experiments of glioblastoma samples with CD-DBCO 2 without the Abs, showed no labelling further demonstrating that Abs-CD 13a is responsible for the labelling observed (see ESI, Fig. S26a †). As a comparison, Invitrogen SiteClick Qdot 625 antibody labeling kit was also used to label anti-GFAP Ab, however anti-GFAP Ab-QD probes tended to precipitate and a stable suspension was only possible for a few minutes aer preparation. Moreover, labelling of brain tumour samples with abfunctionalised-QDs lead to low levels of labelling when compared to our CD-probes and most of the labelling is associated to necrotic tissue. In addition, unfunctionalized acid coated QDs (Qdot™ 625 ITK™ carboxyl quantum dots), as a negative control, showed wide spread non-specic labelling in brain tissue samples. These results further validate the utility of using CDs as the uorescent platform ( Fig. S28 in ESI †).

Conclusions
In summary, we have successfully developed a new class of carbon dot-based uorescence labels that can be "clicked" onto suitably functionalised proteins such as Abs in a chemoselective manner. The two-step strategy relies on the used DBCOfunctionalised CDs and azide-functionalised proteins that can be easily prepared by simple amide conjugation methods from suitably functionalised linkers with control over degree of functionalization. The novel anti-GFAP Abs-CD probes developed here retained their ability to interact with the human GFAP. Moreover, we have demonstrated our novel probes show reliable binding in a range of clinical malignant brain tumour glioblastoma cases, in tissue sections. Overall, this new class of probes offer a promising and versatile strategy in terms of costs, photostability and applicability which can be extended to other Abs and protein systems. This type of cheap and rapid nanoparticle tests have the potential to pave the way for novel strategies to identify the presence of tumour markers such as GFAP in clinical samples to support early diagnosis of brain tumours in primary care. Early diagnosis would potentially improve survival and reduce anxiety in these patients by giving them more surgical and treatment options earlier in the course of their disease.

General experimental
Reagents and solvents were purchased as reagent grade from Sigma Aldrich or ThermoFisher and used without further purication. For column chromatography, silica gel 60 (230-400 mesh, 0.040-0.063 mm) was purchased from Merck and for gel ltration Sephadex G-25 from GE Healthcare. Thin Layer Chromatography (TLC) was performed on aluminium sheets coated with silica gel 60 F254 purchased from Merck. Dialysis was performed with Cole-Parmer Spectra Por Dialysis Tubing, 500-1000 MWCO. Centrifugal spin ltration was performed on Amicon Ultra-0.5 mL purchased from Merck. Carbon dots were prepared using a domestic microwave oven (300 W). NMR spectra were recorded on Bruker AV 400 MHz or AV 500 MHz spectrometers, using the residual solvent peaks as internal reference at 298 K. Chemical shis are reported as parts per million and coupling constants (J) given in Hertz. All the assignments were conrmed by one-and two-dimensional NMR experiments (DEPT, COSY, HSQC). Mass spectra were obtained by the University of Bristol mass spectrometry service using electrospray ionisation (ESI) acquired on a Micromass LCT mass spectrometer or aVG Quattro mass spectrometer and MALDI spectra were acquired on Bruker ultraeXtreme 2 (TOF). Zeta potential analysis was carried out using Malvern Instruments Nano-Z ZEN 2600 and conducted in distilled H 2 O at a concentration of 4 mg mL À1 . BSA was purchased from Ther-moFisher (23209). Polyclonal Rabbit Anti-Glial Fibrillary Acidic Protein was purchased from Agilent Dako (Z033401-2). Recombinant Human GFAP protein, used in the western blot test was purchased from abcam (ab114149). Secondary IRDye® 680RD Goat anti-Rabbit IgG used in the western blot test were purchased by LI-COR. Human brain tissue samples were kindly provided by the Southmead Hospital, (University of Bristol, UK).

DBCO-CDs 2.
To a stirred solution of CDs 3 (18.4 mg) in dry DMF (1.84 mL), HATU (13.4 mg, 0.035 mmol) and DIPEA (6.1 mL, 0.035 mmol) were added and the solution was allowed to stir for further 15 minutes at room temperature. A solution of 9 (9.2 mg, 0.018 mmol) in dry DMF (0.5 mL) was added and the solution was stirred at room temperature for 5 h. H 2 O (0.5 mL) was then added to quench the reaction and the solution was stirred for further 10 minutes at room temperature and concentrated under reduced pressure. The residue was redissolved in aq. 0.1 M NaOH solution (3 mL) and stirred for 1 h at room temperature. The pH was neutralized with the addition of aq. HCl 1 M solution (0.15 mL), diluted with H 2 O (20 mL), washed with Et 2 O (5 Â 10 mL), and the water phase was concentrated under reduced pressure. The residue was puried via 1 KDa cut-off dialysis membrane against water, changing the water bath 3 times over a 24 h period. The puried solution was then freeze-dried furnishing 2 (10.2 mg) as a pale yellow solid. 1

General protein/Abs conjugation procedure
Step 1azide functionalization: to a solution of protein in PBS (100 mL, 36.1 mM for BSA or 100 mL, 11.1 mM Abs), different amounts of compound 1 (0.1 mg mL À1 in DMSO stock solution) from 0.14 to 4.81 mmol for BSA and 0.12-2.41 mmol for Abs, were added, respectively (see Tables S1 and S2 †). The nal solution was mixed in a shaker at 400 rpm for 4 h at room temperature. The product was puried via spin-ltration using 30 kDa or 50 kDa cut-off membrane for BSA or Abs respectively, at 4000 g per 20 minutes. The concentrated protein solution was diluted with 100 mL of PBS and concentrated again; this washing step was repeated two more times to remove unbound linker 1 and byproducts of the reaction, furnishing a concentrated 10a-f or 12a-d for BSA and Abs derivatives respectively.
Step 2 -CD-conjugation: the concentrated 10a-f or 12ad solution prepared in Step 1 was diluted with 100 mL of a PBS solution containing DBCO-CD 2(2 mg mL À1 ) mixed in a shaker at 400 rpm for 16 h at room temperature. The product was puried via spin-ltration using 30 kDa or 50 kDa cut-off membrane for BSA or the Abs respectively, at 4000 g per 20 minutes. The concentrated protein solution was diluted with 100 mL of PBS and concentrated again; this washing step was repeated three more times to remove the excess of 2(4 washing steps were judged enough to remove the excess of 2since no uorescence was detected by the naked eye in the washing solution passing through the membrane under UV lamp in the last wash), furnishing a concentrated 11a-f or 13a-d solution.

Gel electrophoresis
SDS-PAGE on 4-12% NuPage gels (Life Technologies) was performed for labelled and unlabelled protein/Abs samples. Loading dye was added to each sample and heated at 100 C for 5 minutes. SeeBlue™ (ThermoFisher Scientic) was used as a ladder. Lanes were loaded at similar protein concentrations and the gel was run with MES buffer at 180 V for 40 minutes. The images were acquired aer staining with Pageblue (Ther-moFisher Scientic/Pierce) protein staining solution.

Western blot analysis
Recombinant human GFAP protein were run on 12% SDS-PAGE and transferred onto membrane via Trans-Blot Turbo Transfer System (BIO-RAD). Membrane was incubated in blocking buffer (PBS, 0.1% Tween and 2% milk) for one hour and then incubated in primary anti-GFAP antibodies or anti-GFAP-CD conjugates 13a (results and discussion) overnight in the cold room. Membranes were washed and then incubated in secondary antibodies for one hour. The membranes were washed and then visualized on a LI-COR Odyssey imaging system.

Tissue staining protocol for GFAP immunouorescence with Abs-CD 13a
The tissue sections were deparaffinized and rehydrated as follows, the sections were incubated in three washes of xylene for 2 min each, followed by two washes of 100%, 95% ethanol for 10 min each. The sections were then washed twice in distilled H 2 O for 5 min each.
The tissue slides were then placed in the microwaveable vessel. Tris-EDTA antigen retrieval buffer (10 mM Tris base, 1 mM EDTA solution, 0.05% Tween 20, pH 9.0) was added and placed inside a dedicated domestic microwave, which was set to full power (950 W) until the solution came to a boil. The solution was boiled for 20 min from this point and le on the bench at room temp to cool for 30 min. The slides were then washed 2 Â 5 min with TBS plus 0.025% Triton X-100 with gentle agitation. The slides were blocked in Superblock buffer (Thermo-sher, ref. 37515) 30 min at room temp. The slides were drained for a few seconds (not rinsed) and wiped around the sections with tissue paper. 400 ml of CDs-conjugated GFAP antibody 13a (1 : 500) were then added per slide and incubated at 4 C overnight. The slides were then rinsed 3 Â 5 min with TBS plus 0.05% Tween20.
4.6.1 Nuclear stain. The slides were equilibrated with 300 ml buffer 2x SSC (0.3 M NaCl, 0.03 M sodium citrate, pH ¼ 7.0) 2 Â 3 min, then 150 ml (500 nM) propidium iodide (Thermosher, cat.no P3566) were added per slide, incubated at 37 C incubator for 5 min. Aerwards, the slides were washed 6 times with buffer 2x SSC 300 ml. The slides were mounted using mounting medium uromount-G and a coverslip was added. Clear nail polish was added to seal the edges around the coverslip.

Confocal microscopy
Optical microscope images were acquired on a Leica DMIL Led Fluo microscope. Confocal microscope images were acquired on a Leica DMi8 inverted epiuorescence microscope using 405 nm and tuneable white light lasers and 63Â (NA 1.4) objective at the Wolfson Imaging facility at the University of Bristol. The images were analysed using Fiji (ImageJ) soware.