Biomaterials 35 (2014) 2987e2998

Contents lists available at ScienceDirect

Biomaterials journal homepage: www.elsevier.com/locate/biomaterials

Bacterial imaging with photostable upconversion fluorescent nanoparticles Li Ching Ong a, Lei Yin Ang b, Sylvie Alonso a, b, **, Yong Zhang a, c, * a Graduate School for Integrative Sciences and Engineering, National University of Singapore, Centre for Life Sciences (CeLS), #05-01, 28 Medical Drive, Singapore 117456, Singapore b Department of Microbiology, Immunology Program, National University of Singapore, Centre for Life Sciences (CeLS), #03-05, 28 Medical Drive, Singapore 117456, Singapore c Department of Biomedical Engineering, Faculty of Engineering, National University of Singapore, Block EA #03-12, 9 Engineering Drive 1, Singapore 117575, Singapore

a r t i c l e i n f o

a b s t r a c t

Article history: Received 18 November 2013 Accepted 19 December 2013 Available online 10 January 2014

Autofluorescence, photodamage and photobleaching are often encountered when using downconverting fluorophores and fluorescent proteins for bacteria labeling. These caveats represent a serious limitation when trying to map bacteria dissemination for prolonged periods. Upconversion nanoparticles (UCNs), which are able to convert low energy near-infrared (NIR) excitation light into higher energy visible or NIR light, can address these limitations. These particles’ unique optical properties translate into attractive advantages of minimal autofluorescence, reduced photodamage, deeper tissue penetration and prolonged photostability. Here, we report a UCN-based bacteria labeling strategy using Escherichia coli as prototypic bacteria. A comparative analysis highlighted the superior photostability of UCN-labeled bacteria over green fluorescent protein-expressing bacteria. Infection study of UCN-labeled bacteria in dendritic cells indicated co-localization of the UCN signal with bacterial position for up to 6 h postinfection. Furthermore, long-term monitoring of the same infected cells demonstrated the potential to utilize photostable UCN-based imaging for bacterial trafficking purposes. Ó 2013 Elsevier Ltd. All rights reserved.

Keywords: Bacteria labeling Upconversion nanoparticles Fluorescence imaging Escherichia coli

1. Introduction The ability for a microbial pathogen to disseminate from its point of entry to other target organs into its host represents a critical aspect of pathogenesis. Importance of dissemination has been illustrated in several disease examples, where the extent of pathogen dissemination to draining lymph nodes or other organs has been directly correlated with disease severity and manifestations [1,2]. Therefore, mapping the trafficking profile of a pathogen is of great interest to gain further insights into hostepathogen interactions and the mechanisms involved in a pathogen’s ability to cause disease [3]. Furthermore, for attenuated bacteria and other non-pathogenic bacteria which may be used as vaccine candidates

* Corresponding author. Department of Biomedical Engineering, Faculty of Engineering, National University of Singapore, Block EA #03-12, 9 Engineering Drive 1, Singapore 117575, Singapore. Tel.: þ65 6516 4871; fax: þ65 6872 3069. ** Corresponding author. Department of Microbiology, Yong Loo Lin School of Medicine, Immunology Program, Life Science Institute, National University of Singapore, Centre for Life Sciences (CeLS) building, #03-05, 28 Medical Drive, Singapore 117456, Singapore. Tel.: þ65 6516 3673; fax: þ65 6778 2684. E-mail addresses: [email protected] (S. Alonso), [email protected] (Y. Zhang). 0142-9612/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.biomaterials.2013.12.060

or carriers, information of their biodistribution, persistence and clearance in the host is also essential for establishing their safety and efficacy profiles, and for a rational design of bacteria-based vaccine delivery systems [4]. Current approaches of bacteria labeling generally involve either tagging the bacteria with custom designed fluorescent probes [5e 8] or generating recombinant bacteria that express luciferase or fluorescent proteins [3]. Luciferase-expressing recombinant bacteria require the constant presence of its substrate that may be toxic at high concentrations. Furthermore, signal production is highly dependent on bacterial metabolic state and oxygen content in the environment [3,9] On the other hand, when using organic dyes as probes or recombinant bacteria expressing fluorescent proteins, photobleaching, photodamage and autofluorescence issues are commonly encountered. Despite efforts to improve photostability of fluorescent dyes and fluorescent proteins by limiting power exposure [10] and reducing exposure to quenchers [11,12], the problems are not completely resolved, and these approaches still do not allow the study of bacterial dissemination in vivo which requires long and/or repeated exposure times. There is thus a need for alternative bacterial labels that can overcome the current limitations.

2988

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

Though efforts have been made to utilize semiconductor-based quantum dots (QDs) as alternative bacterial labels [13e15], the potential cytoxicity risk associated with their heavy metal components such as cadmium remains a major concern for the use of QDs in biological studies [16]. In addition, since the fluorescence from QDs relies on energy downconversion, background autofluorescence remains a concern as shorter wavelength light is used to excite the nanoparticles. Autofluorescence interferes with signal to background ratio and adversely affects detection limits. This has presented a major hurdle for monitoring bacterial trafficking in vivo as usually low number of bacteria are tracked. Upconversion nanoparticles (UCNs) represent an emerging class of luminescent nanomaterial, which have been shown to be free of autofluorescence in vivo and with a superior detection limit when compared to QDs [17]. These advantages are conferred by the unique ability of UCNs to upconvert light by sequential multiphoton light absorption, where lower energy near-infrared (NIR) excitation light is converted into emission light of higher energy [18e20] The near-infrared excitation wavelength of UCNs lies close to the optical window, a spectral region from 650 nm to 950 nm in which biological materials have low absorption and scattering coefficients, thus leading to negligible autofluorescence and increased tissue penetration [21]. Hence, UCN-based labeling technology has become appealing as it allows for deeper tissue penetration, low autofluorescence and minimal photodamage (as depicted in Fig. 1). In addition, the extreme photostability of these nanoparticles has also been well-established [19]. This implies that monitoring of UCN-labeled entities such as mammalian cells or bacteria for example can be carried out without any decrease of the signal intensity over long periods of time. Last but not least, UCNs have been associated with low cytotoxicity in various in vitro and in vivo settings, supporting their suitability for biological applications [18,19]. As such, UCN labeling has been successfully

reported for various mammalian systems both in vitro and in vivo [18e20,22]. However, the use of UCNs for bacteria labeling has never been reported before. The successes of labeling and imaging of mammalian cells strongly suggest the feasibility of extending UCN labeling to the bacteria kingdom and help to overcome the limitations currently faced using fluorescent dyes and proteins for bacterial trafficking. We report here the successful labeling of prototypic Escherichia coli (E. coli) bacteria with UCNs. The resulting UCN-labeled bacteria exhibited strong UCN luminescence upon NIR excitation. Superior photostability of the UCN-labeled bacteria was demonstrated by comparing its photobleaching profile with that of green fluorescent protein (GFP)-expressing E. coli. Monitoring of UCNlabeled bacteria upon dendritic cell infection was also performed and showed good correlation between UCN signal and bacteria positioning. 2. Materials and methods 2.1. Synthesis of antibody-conjugated-UCNs 2.1.1. Synthesis of citrate-UCN All chemicals used here were purchased from SigmaeAldrich and used without further purification. Oleic acid-capped NaYF4:Yb,Er UCNs were first synthesized using a previously reported reaction scheme [23]. One mmole of reaction salts (YCl3, YbCl3 and ErCl3) in the stoichiometric ratio were dissolved in 6 ml of oleic acid and 15 ml of octadecene by heating the mixture at 160  C for about 30 min until a homogenous solution was formed. After cooling the solution to room temperature, 2.5 mmol of sodium hydroxide and 4 mmol of ammonium fluoride (dispersed in 10 ml of methanol) were added. The mixture was then slowly heated for methanol removal and also degassed at 100  C for 10 min. After which, the mixture was maintained at 300  C under an argon atmosphere for 1 h. The nanoparticles were purified by acetone precipitation and the other insoluble impurities were removed by low speed centrifugation. Purified oleic acid-capped-UCNs were stored in cyclohexane until further use for citrate modification. Citrate modification of the OA-UCNs was performed via a ligand exchange process as described previously [24,25]. Briefly, a mixture of 2 mmol sodium citrate

Fig. 1. Schematic diagram illustrating the bacteria labeling strategy to obtain UCN-labeled E. coli that can be used for various in vitro and in vivo studies, while having the advantages of low autofluorescence, deeper penetration, minimal photodamage and photostability.

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998 in 15 ml of diethylene glycol was first heated it to 110  C for 30 min under an argon atmosphere. Oleic acid-capped-UCNs (10 mg) dispersed in cyclohexane and toluene were then added into the mixture and the reaction was heated to 160  C for evaporation of cyclohexane and toluene. Upon complete evaporation, the reaction was further maintained at 160  C for 3 h. Citrate-UCNs were then collected by centrifugation, washed with ethanol and deionized water and finally dispersed and stored in deionized water. 2.1.2. Antibody conjugation to citrate-UCNs Citrate-UCNs (1 mg/ml in water) were activated with 10 mM N-(3Dimethylaminopropyl)-N0 -ethylcarbodiimide hydrochloride (EDC, Sigmae Aldrich, Cat. No. E7750) and 26 mM N-hydroxysuccinimide (NHS, SigmaeAldrich, Cat. No. 130672) for about 3 h at room temperature. The activated citrate-UCNs were purified by centrifuging at 9350 g for 15 min and re-dispersed in ultrapure water. 100 mg of anti-E. coli antibody (Abd Serotec, Cat. No. 4329-4906) were incubated with 1 mg of activated particles. After overnight shaking at 4  C, unbound antibodies were removed from the antibody-UCNs by centrifugation (6000 g, 10 min at 4  C) and washing the particles once with ultrapure water. The antibody-UCNs were then re-dispersed in ultrapure water. Control-UCNs were prepared as described above with the exception that an irrelevant antibody (Bovine IgG, Cat. No. 001-0102, Rockland Immunochemicals Inc.) was added instead of the anti-E. coli antibody.

2989

electron microscopy (TEM) viewing of the bacterial sections was performed with TEM Philips 208s. 2.4. Photostability comparison of UCN and GFP signals UCN-labeled E. coli resuspended in culture media were constantly irradiated under the fluorescence microscope (Nikon 80i upright fluorescence microscope, fitted with cytoviva module and 980 nm laser, Nikon Plan Fluor oil-immersion 100 objective, adjustable iris, NA fixed at 1.3) with either blue light for GFP excitation (power density of w13.5 mW/cm2, Chroma 31001filter set, excitation 480/30 nm and emission 535/40 nm) or NIR laser for UCN excitation (power density of w3 W/cm2 and emission filter ET850sp-2p-1500 625/450 nm). Timelapsed images were obtained every 15 s over the 30 min period. Photobleaching curves were plotted by normalizing the corrected intensity (Equation (1)) of the green channel (brightest channel for both UCN and GFP) to that at time zero. Corrected Total Intensity ¼ Integrated Density  ðArea  Mean of Background MeanÞ

(1)

Equation 1: Calculation of corrected total intensity for image intensity measurements. Mean of background mean refers to the average pixel intensity of three randomly chosen regions of the background in the same image.

2.2. Particles characterization

2.5. Stability of association of UCNs with E. coli

Particles’ size and polydispersity index (PdI) were determined by dynamic light scattering using Malvern Nano-ZS instrument. Visible luminescence spectra (from 400 to 700 mm) of the particles were acquired with a SpectroPro 2150i spectrophotometer equipped with 980 nm diode laser excitation (600 mW). Transmission electron microscope (TEM) images of the citrate-UCNs were recorded on a field emission transmission electron microscope JEOL-JEM 2010F. Fourier transforminfrared spectroscopy (FT-IR) was obtained by IR-Prestige (Shimadzu) in the attenuated total reflectance mode with the air-dried concentrated samples of particles and antibody. Protein content of the antibody-UCNs was estimated using Bradford microassay (Cat. No. 500-0006, Bio-rad Laboratories). The standard curve was obtained by using bovine IgG as the standard protein.

UCN-labeled bacteria were pelleted and resuspended in the respective solutions (DMEM media (with and without serum), saline (0.9% sodium chloride), LB broth and PBS) and incubated at 37  C for about 1 h. After incubation, the bacteria were pelleted by centrifuging at 840 g for 5 min. Visible emission spectra of the bacteria and the supernatants upon 980 nm NIR excitation were measured with a SpectroPro 2150i spectrophotometer equipped with 980 nm diode laser under the same power (w650 mW) and detector settings. Luminescence intensities of the various suspensions were obtained from OriginPro by summing up the total area under the spectra curves. Percentages of UCN signal associated with the bacteria were calculated by normalizing the amount of UCN signals measured for the bacterial pellet with respect to the total UCN signals for the sample (sum of signals from the pellet and supernatant).

2.3. Bacteria labeling 2.3.1. Bacterial culture and preparation Green Fluorescent Protein (GFP) expressing E. coli (GFP-E. coli) were prepared by transforming HST08 E. coli strain with pZsGreen Vector (Clonetech Labs). After overnight growth in LB broth (Sigma Aldrich, Cat. No. 51208), GFP-E. coli were pelleted by centrifuging at 6000 g for 5 min and washed twice with deionized water prior to bacterial labeling.

2.6. Cell culture DC2.4 cells, an immature murine dendritic cell line, (a kind gift from Dr Wong Siew Heng) were cultured in complete DMEM media (Gibco Cat. No. 11965-092, supplemented with 10% fetal bovine serum (FBS)) in a humidified 37  C incubator with 5% CO2. 2.7. Infection study in dendritic cells

2.3.2. Bacteria labeling with antibody-UCNs Bacterial labeling was carried out as depicted by Fig. 1. 50 ml of the antibodyUCNs (1 mg/ml) or control-UCNs (1 mg/ml) were mixed with the approximately 106 colony forming unit (c.f.u.) of E. coli bacterial pellet and incubated for 3 h at 37  C with agitation. After which, the bacterial suspensions were pelleted by centrifuging at 840 g for 5 min and washed thrice with ultrapure water to remove unbound nanoparticles. 2.3.3. Spectra measurement Visible spectra of the bacterial suspensions (from 400 to 700 nm) were measured with a SpectroPro 2150i spectrophotometer equipped with 980 nm diode laser excitation under the same excitation power (w650 mW) and detector settings. Total intensities of bacterial solution were obtained from by summing up the area under the spectra curves in OriginPro 9 software. 2.3.4. Imaging of bacteria by fluorescence and electron microscopy The bacteria samples were wet-mounted and viewed under a Nikon 80i upright fluorescence microscope (fitted with cytoviva module) to acquire the GFP (Chroma 31001filter set, excitation 480/30 nm and emission 535/40 nm) and UCN signals (980 nm laser and emission filter ET850sp-2p-1500 625/450 nm) with 100 objective (Nikon Plan Fluor oil-immersion 100 objective, adjustable iris, NA fixed at 1.3). The UCN images for intensity measurement were acquired using constant NIR laser settings (w3 W/cm2) and constant camera settings. For a clearer presentation of the UCN signals, the raw RGB images of the UCN signals were split into their respective channels and the strongest green signal of the UCN images were pseudocoloured red in ImageJ (without brightness or contrast enhancements so as to illustrate the inherent intensity differences between the different samples). For electron microscopy, untreated and UCN-labeled E. coli were fixed overnight at 4  C with 4% buffered glutaraldehye. Osmium tetraoxide post-fixation step were included to counter-stain the bacteria. For field emission scanning electron microscopy (FESEM), the bacteria were sputtered with carbon and viewed using Philip XL30 FEG SEM (accelerating voltage of 20.0 kV and a 60:40 mix of secondary electron and back-scattered electron mode). Transmission

DC2.4 dendritic cells were grown onto plasma-treated coverslips (placed in 24well plate) at a concentration of 7  105 cells per well about 24 h prior to the infection. Immediately prior to infection, the cells were washed once with DMEM media without serum (DMEM infection media) and the incubation media replaced with this DMEM infection media. About 106 c.f.u of UCN-labeled bacteria were coincubated with the cells at 37  C for 1 h. The cells were then washed thrice with complete DMEM media to remove the free bacteria. The washed cells were replaced with fresh complete DMEM media and further incubated at 37  C and in 5% CO2. About 10 min before the stipulated time points, the cells were stained with a plasma membrane dye (5 mg/ml of CellMaskÔ Orange plasma membrane stain, Invitrogen Cat. No. C10045) at 37  C. At the indicated time points, the cells were washed thrice with complete DMEM media and once with phosphate buffered saline to remove the excess membrane dye and subsequently fixed with 4% paraformaldehyde. After fixation, the cell nuclei were also stained with 40 ,6Diamidino-2-phenylindole dihydrochloride (DAPI, Sigma Aldrich, Cat. No. D8417). The fixed cells were then histo-mounted and viewed with the Nikon 80i upright fluorescence microscope (fitted with cytoviva module and 980 nm laser) at either 40 (Nikon Plan Fluor, NA 0.75) or 100 magnification (Nikon Plan Fluor, adjustable iris, NA fixed at 1.3). 2.8. Long-term imaging of UCN-labeled E. coli in dendritic cells Infected dendritic cells were prepared similarly as in previous section (Infection Study in Dendritic Cells). After washing the cells thrice with complete DMEM media to remove the free bacteria, the cells were stained for their nuclei and plasma membrane with Hoechst 33342 (2 drops/ml, Invitrogen, Cat No. R37605) and 5 mg/ ml of CellMaskÔ Orange plasma membrane stain for 10 min (Invitrogen, Cat No. C10045). The cells were then washed thrice with complete DMEM media to remove the excess dyes and were subsequently incubated with complete CO2-independent media (Gibco, Cat No. 18045-088, supplemented with 10% FBS and L-glutamine). The cells were then equilibrated at 37  C for about half an hour on inverted confocal microscope (Nikon Eclipse TE2000-U), before starting the long term monitoring of the UCN signal for about 6 h with the 100 objective (NA 0.90). Widefield fluorescence images of the plasma membrane and bacterial GFP signal

2990

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

Fig. 2. Properties of citrate-UCNs and antibody-UCNs. (A) DLS size graph of (i) citrate-UCN and (ii) antibody-UCNs. Insets are the TEM images of the respective particles. (B) Visible emission spectra (lexcitation ¼ 980 nm) of citrate-UCNs (solid line) and antibody-UCNs (dotted line). (C) FTIR spectra of (i) citrate-UCN, (ii) pure antibody and (iii) antibody-UCNs.

were taken once at the start of the acquisition while that of the nuclei were taken approximately every half an hour (after every 2e3 exposures of UCN signal). Five zsections (0.7 mm apart) were taken for the UCN signal and the average z-projection for each time point was merged with the nuclei’s Hoechst signal.

Average projections of the raw z-sections were used for the measurement of the luminescence intensity. Corrected total intensity of the average z-projection for each time point was calculated (Equation (1)) and normalized with respect to the calculated value at the first exposure to obtain the normalized intensity.

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

3. Results 3.1. Synthesis and characterization of UCNs Bacterial-specific antibodies were conjugated onto the surface of citrate-UCNs to yield bacteria targeting UCNs. Among the different recognition elements, the widely used antibodies were chosen here as they are known to have high selectivity and sensitivity. In addition, there are numerous well-established chemistries available for the conjugation of antibodies to nanoparticles [26]. Prior to conjugation, the size quality of the precursor citrate-UCNs was first validated by dynamic light scattering (DLS) and transmission electron microscopy (TEM). The presence of a single peak in the DLS size graph and low polydispersity index (PdI) showed that the citrate-UCNs were of narrow size distribution (Fig. 2A(i)), while TEM revealed that the particles had diameters of about 20 nm (Insets in Fig. 2A). After antibody functionalization of the citrateUCNs to yield the bacteria-targeting antibody-UCNs, the average diameter as determined by TEM and good size quality (low PdI) of the nanoparticles was also maintained (Fig. 2A (ii))). The discrepancy in particles’ size measured by DLS and TEM is due to the fact that DLS calculates the particles’ hydrodynamic diameter based on their Brownian diffusion rate while TEM visualizes the UCNs based on their electron diffraction and do not detect the inherently electron-transparent biomolecules such as antibodies. The presence of the antibody on antibody-UCNs most probably retards the particles’ Brownian motion rates, thereby resulting in a much larger hydrodynamic diameter measured by DLS [27]. The particles’ luminescence properties were also retained as both the antibodyUCNs and citrate-UCNs showed similar characteristic emission peaks at about 410, 526, 544 and 658 nm upon NIR excitation (Fig. 2B). The presence of antibodies on the antibody-UCNs was then confirmed by Fourier transform-infrared spectroscopy (FT-IR) (Fig. 2C). In the region from 1500 to 1600 cm1, the citrate-UCN had a single broad peak at 1635 cm1, corresponding to the asymmetric stretching mode of the citrate ligands’ carboxyl groups. On the other hand, the pure antibody and antibody-UCNs exhibited two sharp peaks at about 1640 cm1 and 1540 cm1, which can be assigned to the distinct amide I and amide II vibration modes of proteins. Gel electrophoresis of the antibody-UCNs under denaturing and reducing conditions only released the light chain fragment of the IgG antibody (25 kDa) (Figure S1), suggesting that the strong covalent linkages between the antibodies and the UCNs occurred primarily via the heavy chain fragment (50 kDa). In addition, Bradford protein assay of the antibody-UCNs estimated the antibody content to be about 40 mg/mg particles. 3.2. Bacterial labeling GFP-expressing E. coli bacteria were incubated with antibodyUCNs or with control-UCNs which were functionalized with an irrelevant antibody. The bacteria co-incubated with antibody-UCNs (Fig. 3B) displayed strong nanoparticles’ luminescence, while those co-incubated with the control-UCNs (Fig. 3A) only displayed weak background signals. In addition, uniform coating of the bacteria with the antibody-UCNs was achieved as evidenced by the UCN signal delineating the rod shape morphology of bacteria and colocalizing with the bacterial GFP signal (Fig. 3B). Bacteria labeling efficiency of the entire bacterial population was assessed by measurement of UCN signal intensities of the bacterial suspensions. Representative visible spectrum of the bacteria labeled with antibody-UCNs displayed peaks of much higher intensities as compared to that of bacteria labeled with control-UCNs (Fig. 3C (i)). Upon normalizing the intensity, the total UCN signal intensity of the

2991

bacterial suspension labeled with antibody-UCNs was significantly higher (p < 0.05) than that of the bacteria labeled with controlUCNs (Fig. 3C (ii)). Taken together, the results demonstrated the targeting ability of the antibody-UCNs and the high efficiency of this labeling strategy. Electron microscopy (EM) was also performed to determine if UCNs remain at the bacterial surface or have been internalized. Under field emission scanning EM (FESEM), UCN-labeled E. coli exhibited a rough surface morphology which is in sharp contrast to the smooth surface of the unlabeled bacteria (Fig. 4A). To confirm that the rough surface phenotype observed by FESEM is due to the presence of the UCNs, cross sections were examined by TEM. Electron-dense dark spheres of about 20 nm in diameter, which corresponds to the size of the UCNs, decorated the bacterial outer surface (Fig. 4B). No signal was observed inside the cytoplasm of the UCN-labeled bacteria indicating that the UCNs have not been internalized by the bacteria. In addition, no signal was observed with unlabeled bacteria. Electron microscopy approaches thus confirmed the presence of UCNs, uniformly distributed on the bacterial surface. 3.3. Photostability comparison of UCN and GFP signal Photobleaching is a common problem encountered when using conventional fluorophores and fluorescent proteins. This problem is more pronounced when monitoring is performed for extended periods of time, during kinetic and trafficking studies for example. As mentioned earlier, the unique photophysical properties of UCNs can help circumvent the photobleaching issues. To test this hypothesis, a comparative analysis of the photostability of the GFP and UCN signals was performed. Continuous irradiation of GFPexpressing UCN-labeled bacteria with blue light resulted in an exponential decrease of the GFP signal over time with a half-life of 75 s (Fig. 5A). In contrast, exceptional photostability of the UCN signal was observed as the maintenance of a constant and optimal intensity during the entire monitoring period. The drastic difference in photostability of GFP and UCN signals is also evident from the time-lapsed images of the labeled bacteria (Fig. 5B). The GFP signal (represented in green) rapidly faded within the first 5 min of constant irradiation and was no longer detectable after 10 min. In contrast, the UCN signal (represented in red) was clearly visible throughout the entire observation period with no noticeable reduction in signal intensity. 3.4. Stability of the attached UCN signals in different solutions Since UCNs are attached to the bacteria surface via strong but non-covalent antibodyeantigen interactions, there is a possibility that the nanoparticles may dissociate from the bacteria under certain culture conditions which may thus impair bacteria detection and tracking. To assess this eventuality, UCN-labeled E. coli bacteria were incubated in DMEM media (with and without serum), saline (0.9% sodium chloride), LB broth and PBS. These different incubation media are commonly used for in vitro infection assays and growth kinetic studies, respectively. After incubating the labeled bacteria in the respective solutions, dissociation of UCNs from bacteria was estimated by measuring the relative amounts of UCN luminescence in the supernatant and bacterial pellet after centrifugation. Microscopic examination of the pelleted bacteria was also carried out to verify that the UCN signal detected in the pellet were indeed due to those particles that remained associated to the bacteria (Figure S2). Fig. 6 summarizes the relative percentage of bacterial-associated UCN luminescence after the incubation in the various solutions. The association of UCNs to the bacteria was thus found to be

2992

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

Fig. 3. Bacterial targeting ability of antibody-UCNs. Images (pseudo-colored) of E. coli (labeled with (A) control-UCNs and (B) antibody-UCNs. Zoomed in view of individual bacteria (highlighted in white boxes) are in panel (ii). (C) (i) Representative visible emission spectra (lexcitation ¼ 980 nm) of bacterial suspensions that have labeled with control-UCNs (solid line) and antibody-UCNs (dotted line). (ii) Column plot of total fluorescence intensities (lexcitation ¼ 980 nm) of entire bacterial suspension (normalized to that of the bacteria suspension labeled with control-UCNs). Error bar represents standard error of three independent experiments (*p < 0.05 based on two sample t-test).

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

2993

Fig. 4. Surface morphology of unlabeled E. coli and UCN-labeled E. coli observed by (A) FESEM at (i) low and (ii) high magnification. (B) Cross-sectional view of unlabeled E. coli and UCN-labeled E. coli observed by TEM at (i) low and (ii) high magnification.

stable in DMEM (with or without serum), saline and PBS with more than 90% of the total UCN signal found in the bacteria pellet fraction. In contrast, incubation in LB broth resulted in significant dissociation of the UCNs from the bacteria with only about onethird of the total UCNs signal detected in the bacterial pellet fraction.

3.5. Colocalization of UCN signal with bacteria during infection of dendritic cells In vitro infection assays are widely employed to study hoste pathogen interactions and to determine the intracellular localization of the bacterium upon phagocytosis. For these, efficient

2994

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

Fig. 5. Photostability comparison of UCN and GFP signals upon continuous irradiation. (A) Normalized fluorescence intensity (with respect to the fluorescence intensity at time zero) plots of GFP (square) and UCN (circle) signals of continuously irradiated UCN-labeled E. coli. Error bars represent standard deviation of four independent experiments. (B) Time-lapsed images of UCN-labeled E. coli. Total irradiation time is indicated at top left corner of each image in minutes:seconds format.

tracking of the bacterium is necessary. To evaluate the usefulness of UCN bacteria labeling for intracellular tracking, dendritic cells were infected with GFP-expressing UCN-labeled bacteria. At specific time points up to 6 h post infection, the cells were fixed and stained for their nuclei and plasma membrane in order to mark the cell positions and boundaries. Respective fluorescence images of the

dendritic cells’ nuclei (blue), cell plasma membrane (gray), bacteria (GFP signal, green) and UCNs (red) at different time points are presented in Fig. 7. Examination of the general cell population at lower magnification showed that the UCN signal co-localized perfectly with the GFP signal for the entire observation period. This co-localization is more evident from the merged images of

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

Fig. 6. Stability of UCN association to bacteria upon incubation in different physiological relevant solution. Error bars represent standard deviation of three independent experiments.

both GFP and UCN signals, in which the two signals were observed to overlap with each other. In addition, for the entire 6 h period, UCN signals continued to delineate the bacterial rod shape morphology as seen from the higher magnification fluorescence images (insets in Fig. 7). 3.6. Long term monitoring of UCN-labeled E. coli in dendritic cells The data from the previous section indicated that bacteria UCN labeling is as effective as GFP signal to inform on the bacteria positioning inside the mammalian cells. It is also essential to investigate if there is a decrease in UCN signal over time due to dissociation of the UCNs from the bacteria within intracellular environment, which may then adversely affect the detection limit. To investigate if there is such dissociation and to demonstrate the application of the UCN-labeled bacteria for photostable imaging, long term monitoring of the same set of dendritic cells infected with UCN-labeled bacteria was performed. Labeled bacteria within dendritic cells were tracked via the UCN signal while the relative cell positions were periodically checked by the Hoechst signal from the stained nuclei. As shown in Fig. 8A, information on relative bacterial-cell nuclei location can be determined throughout the entire observation period by superimposing the UCN and Hoechst signals. Quantification of the UCN signal intensity for this same set of bacteria within the dendritic cells showed that the signal intensity remained relatively constant throughout the entire observation period (Fig. 8B). The slight fluctuations of the measured intensity observed can be attributed to the minute movement of the bacteria and cells. This meant that certain regions of the bacteria may be out of focus at a specific time point, thereby resulting in some small variations in the amount of luminescence signal collected. Overall, the results imply that there was no significant dissociation of UCN particles from the bacteria for up to 6 h post infection and also illustrate the possibility of using the photostable UCN signal in long term monitoring studies. 4. Discussion In this paper, we report the synthesis and use of antibodyconjugated UCNs that specifically and uniformly labeled

2995

prototypic E. coli bacteria. The UCN-labeled bacteria exhibited strong and detectable UCN luminescence. Electron microscopy showed that the UCNs particles were not internalized by the bacteria and instead were uniformly distributed at the bacterial surface. The size of the UCNs (20 nm in diameter) likely prevented from passive diffusion through the bacterial cell wall since the size limit for this route lies in the range of several nanometers [28]. In addition, while it may be possible for nanoparticles up to about 100 nm to be specifically uptaken into bacteria via some active transporter mechanism [29,30], this approach necessitates the nanoparticles surface to be modified with specific ligands that target these transporters. In order to compare UCN versus GFP signals, we made use of GFP-expressing E. coli bacteria that we labeled with the antibodyconjugated UCNs. By simply adjusting the excitation and emission wavelength windows, we were thus able to visualize and monitor either signal. Our comparative analysis between these two signals has clearly demonstrated the superior photostability of UCN. Consistent with previous reports [11,12,31], GFP displayed a typical photobleaching half-life of 75 s and the signal specificity was lost completely after 10 min. Photobleaching phenomenon is attributed to a photochemical reaction, termed as oxidative redding. In the presence of electron acceptors, GFP acts as electron donor and is converted into a red fluorescent energy state and eventually into a photobleached state [32]. While efforts have been made to reduce photobleaching by limiting the exposure of GFP to electron acceptors through the omission of certain cell culture media components [11,12], the reported photobleaching half-life of GFP still remains within the range of several minutes. In contrast, our results showed that bacteria-associated UCNs exhibited constant signal intensity throughout 30 min of continuous excitation. These findings are in agreement with previous studies which reported that UCNs signal did not photobleach upon several hours of continuous excitation [33]. The superior photostability demonstrated for UCNlabeled bacteria thus represents a unique and very attractive advantage over fluorescent tags thereby allowing for extended monitoring without compromising the signal intensity. Antigeneantibody interactions are of reversible non-covalent nature and include interactions such as electrostatic forces, hydrogen bonds, Van der Waals forces and hydrophobic forces [34]. We thus anticipated that the association between antibodyconjugated UCNs and the bacterial surface may be disrupted in the presence of high salts, extreme pH and detergents. The stability of the UCNs association to the bacterial surface was thus determined under different culture conditions. UCNs remained associated to the bacteria in physiological solutions including PBS, saline and standard culture media for mammalian cells. However, incubation of UCN-labeled bacteria in LB broth, which is the standard growth media for E. coli, led to substantial dissociation of the particles from the bacteria. Dissociation could be actually observed immediately upon resuspension of the bacteria into the medium (data not shown). This observation therefore indicates that the dissociation is not due to bacterial growth which would result in dispersion of the UCNs between the daughter cells thereby leading to a progressive decrease of the UCN signal intensity per bacterium. LB broth contains tryptone, yeast extract and sodium chloride, however the exact composition of this culture medium is not clearly defined. Tryptone is a rich source of peptides and amino acids, and serves as a source of essential amino acids for bacterial growth. Competition between the high concentration of amino acids from tryptone and the antibodies for non-covalent interactions with the bacterial surface may be involved in UCN dissociation from the bacterial surface. This hypothesis remains however to be experimentally investigated. Modifications can be made to the existing systems to strengthen the UCN-bacteria

2996

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

Fig. 7. In vitro colocalization behavior of UCN signals with bacterial position upon uptake of UCN-labeled E. coli into dendritic cells. Insets show the respective signals at higher magnification. (Legend: Blue e Cell Nuclei (stained with DAPI), GrayGray e Plasma Membrane (stained with CellMaskÔ Orange), Green e Bacterial GFP signal and Red e UCN signal). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

binding if the disruption of the antigeneantibody interaction was found to be the main reason for the dissociation. Customizable nucleic acid-based aptamers may be used in replacement of the antibodies to mediate attachment of the UCNs to the bacterial surface. From a general pool of synthetic nucleic acids, application and target-specific aptamers are identified by an in vitro iterative process (systemic evolution of ligands by exponential enrichment, SELEX) comprising of multiple cycles of absorption, recovery of the target-bound nucleic acids and polymerase chain reaction amplification of the bound nucleic acids [35]. Hence, the selection process may be customized such that the absorption step is carried out in LB broth so as to identify desirable aptamers that have specificity towards target bacteria and high binding stability in this bacterial growth media. In vitro infection study of UCN-labeled bacteria in dendritic cells showed that there was co-localization of the UCN signal with GFPexpressing bacteria for a period of up to 6 h post-uptake in these cells. This thus suggests that it is possible to utilize the UCN signal to map bacterial trafficking in dendritic cells and study the fate of

bacteria within dendritic cells for up to 6 h. Pushing further the investigation, we showed that photostable UCN-mediated imaging can be combined with other staining approaches such as nuclei Hoechst-staining, thereby illustrating the potential of UCN-labeling for in vitro imaging applications. In addition, the UCN-based imaging approach may also be applied in vivo as deep tissue penetration of up to about 20 mm using NIR to NIR UCNs has already been reported [22]. As such, tracking UCN-labeled bacteria migration upon subcutaneous injection all the way to the draining lymph nodes can be achieved, while taking advantage of the enhanced photostability, low autofluorescence, minimal photodamage and deeper tissue penetration to perform live imaging over an extended period of time. 5. Conclusion The work described here pioneers the use of UCNs as a new bacteria labeling strategy and opens up an entire new application field for this promising luminescent nanomaterial in bacterial

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

2997

Fig. 8. (A) Long term monitoring of the same set of UCN-labeled E. coli in dendritic cells utilizing the UCN signal (Legend: Blue e Cell Nuclei (stained with Hoechst 33342), Gray e Plasma Membrane (stained with CellMaskÔ Orange), Green e Bacterial GFP signal and Red e UCN signal). Total lapsed time (post-infection) indicated in top left corner in hours:minutes:seconds format. (B) Normalized UCN signal intensity (normalized to the intensity at first exposure) plot of labeled E. coli in dendritic cells over time. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

2998

L.C. Ong et al. / Biomaterials 35 (2014) 2987e2998

trafficking. Antibody-based bacterial targeting UCNs specifically and uniformly labeled protypic gram-negative bacteria, E. coli, allowing the labeled bacteria to display strong and distinct upconversion luminescence. Comparative analysis of this upconversion luminescence from the labeled bacteria and the bacterial expressed fluorescent protein demonstrated the photostability advantage of UCNs. Furthermore, upon infection of dendritic cells in vitro, the UCN signals from the labeled bacteria colocalized and persisted with the bacteria for a 6 h observation period, strongly suggesting the possibility of using the labeled bacteria for longterm trafficking studies. Acknowledgments The authors will like to thank Liu Jinliang for providing the citrate-UCNs used in the initial bacterial labeling experiments, Shashi Ranjan and Lim Swee Yin for their help in these experiments and Josephine Howe Lye Chun from Department of Microbiology, NUS, and the Electron Microscopy Unit from Yong Loo Lin School of Medicine, NUS, for their help in the SEM and TEM experiments. The authors will also like to acknowledge the funding support from A*STAR SBIC and National University of Singapore. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.biomaterials.2013.12.060. References [1] Bar-Haim E, Gat O, Markel G, Cohen H, Shafferman A, Velan B. Interrelationship between dendritic cell trafficking and Francisella tularensis dissemination following airway infection. PLoS Pathog 2008;4:e1000211. [2] Krishnan N, Robertson BD, Thwaites G. The mechanisms and consequences of the extra-pulmonary dissemination of Mycobacterium tuberculosis. Tuberculosis 2010;90:361e6. [3] Andreu N, Zelmer A, Wiles S. Noninvasive biophotonic imaging for studies of infectious disease. FEMS Microbiol Rev 2011;35:360e94. [4] Brennan FR, Dougan G. Non-clinical safety evaluation of novel vaccines and adjuvants: new products, new strategies. Vaccine 2005;23:3210e22. [5] Ning X, Lee S, Wang Z, Kim D, Stubblefield B, Gilbert E, et al. Maltodextrinbased imaging probes detect bacteria in vivo with high sensitivity and specificity. Nat Mater 2011;10:602e7. [6] Leevy WM, Gammon ST, Jiang H, Johnson JR, Maxwell DJ, Jackson EN, et al. Optical imaging of bacterial infection in living mice using a fluorescent nearinfrared molecular probe. J Am Chem Soc 2006;128:16476e7. [7] White AG, Fu N, Leevy WM, Lee JJ, Blasco MA, Smith BD. Optical imaging of bacterial infection in living mice using deep-red fluorescent squaraine rotaxane probes. Bioconjug Chem 2010;21:1297e304. [8] Kong Y, Yao H, Ren H, Subbian S, Cirillo SLG, Sacchettini JC, et al. Imaging tuberculosis with endogenous ß-lactamase reporter enzyme fluorescence in live mice. Proc Natl Acad Sci U S A 2010;107:12239. [9] Andreu N, Zelmer A, Fletcher T, Elkington PT, Ward TH, Ripoll J, et al. Optimisation of bioluminescent reporters for use with mycobacteria. PLoS ONE 2010;5:e10777. [10] Hoebe RA, Van Oven CH, Gadella TWJ, Dhonukshe PB, Van Noorden CJF, Manders EMM. Controlled light-exposure microscopy reduces photobleaching and phototoxicity in fluorescence live-cell imaging. Nat Biotechnol 2007;25: 249e53.

[11] Bogdanov AM, Kudryavtseva EI, Lukyanov KA. Anti-fading media for live cell GFP imaging. PLoS ONE 2012;7:e53004. [12] Bogdanov AM, Bogdanova EA, Chudakov DM, Gorodnicheva TV, Lukyanov S, Lukyanov KA. Cell culture medium affects GFP photostability: a solution. Nat Methods 2009;6:859e60. [13] Leevy WM, Lambert TN, Johnson JR, Morris J, Smith BD. Quantum dot probes for bacteria distinguish Escherichia coli mutants and permit in vivo imaging. Chem Commun 2008:2331e3. [14] Hahn MA, Tabb JS, Krauss TD. Detection of single bacterial pathogens with semiconductor quantum dots. Anal Chem 2005;77:4861e9. [15] Yang L, Li Y. Simultaneous detection of Escherichia coli O157:H7 and Salmonella Typhimurium using quantum dots as fluorescence labels. Analyst 2006;131:394e401. [16] Tsoi KM, Dai Q, Alman BA, Chan WCW. Are quantum dots toxic? Exploring the discrepancy between cell culture and animal studies. Acc Chem Res 2012;46: 662e71. [17] Cheng L, Yang K, Zhang S, Shao M, Lee S, Liu Z. Highly-sensitive multiplexed in vivo imaging using PEGylated upconversion nanoparticles. Nano Res 2010;3:722e32. [18] Wang F, Banerjee D, Liu Y, Chen X, Liu X. Upconversion nanoparticles in biological labeling, imaging, and therapy. Analyst 2010;135:1839e54. [19] Zhou J, Liu Z, Li F. Upconversion nanophosphors for small-animal imaging. Chem Soc Rev 2012;41:1323e49. [20] Cheng L, Wang C, Liu Z. Upconversion nanoparticles and their composite nanostructures for biomedical imaging and cancer therapy. Nanoscale 2013;5: 23e37. [21] Smith AM, Mancini MC, Nie S. Bioimaging: second window for in vivo imaging. Nat Nano 2009;4:710e1. [22] Gu Z, Yan L, Tian G, Li S, Chai Z, Zhao Y. Recent advances in design and fabrication of upconversion nanoparticles and their safe theranostic applications. Adv Mater 2013;25:3758e79. [23] Li Z, Zhang Y. An efficient and user-friendly method for the synthesis of hexagonal-phase NaYF4:Yb, Er/Tm nanocrystals with controllable shape and upconversion fluorescence. Nanotechnology 2008;19. [24] Liu J, Cheng J, Zhang Y. Upconversion nanoparticle based LRET system for sensitive detection of MRSA DNA sequence. Biosens Bioelectron 2013;43: 252e6. [25] Liu Q, Sun Y, Yang T, Feng W, Li C, Li F. Sub-10 nm hexagonal lanthanidedoped NaLuF4 upconversion nanocrystals for sensitive bioimaging in vivo. J Am Chem Soc 2011;133:17122e5. [26] Vikesland PJ, Wigginton KR. Nanomaterial enabled biosensors for pathogen monitoring e a review. Environ Sci Technol 2010;44:3656e69. [27] Hall JB, Dobrovolskaia MA, Patri AK, McNeil SE. Characterization of nanoparticles for therapeutics. Nanomedicine 2007;2:789e803. [28] Demchick P, Koch AL. The permeability of the wall fabric of Escherichia coli and Bacillus subtilis. J Bacteriol 1996;178:768e73. [29] Kloepfer JA, Mielke RE, Nadeau JL. Uptake of CdSe and CdSe/ZnS quantum dots into bacteria via purine-dependent mechanisms. Appl Environ Microbiol 2005;71:2548e57. [30] Jayawardena HSN, Jayawardana KW, Chen X, Yan M. Maltoheptaose promotes nanoparticle internalization by Escherichia coli. Chem Commun 2013;49: 3034e6. [31] Zhang C, Konopka JB. A photostable green fluorescent protein variant for analysis of protein localization in Candida albicans. Eukaryot Cell 2010;9:224e 6. [32] Bogdanov AM, Mishin AS, Yampolsky IV, Belousov VV, Chudakov DM, Subach FV, et al. Green fluorescent proteins are light-induced electron donors. Nat Chem Biol 2009;5:459e61. [33] Idris NM, Li Z, Ye L, Wei Sim EK, Mahendran R, Ho PC-L, et al. Tracking transplanted cells in live animal using upconversion fluorescent nanoparticles. Biomaterials 2009;30:5104e13. [34] Janeway CA, Travers P, Walport M, Shlomchik MJ. The interaction of the antibody molecule with specific antigen. Immunobiology: the immune system in health and disease. 5th ed. New York: Garland Science; 2001. [35] Mairal T, Cengiz Özalp V, Lozano Sánchez P, Mir M, Katakis I, O’Sullivan C. Aptamers: molecular tools for analytical applications. Anal Bioanal Chem 2008;390:989e1007.