Optimisation and Use of Humanised RBL NF-AT-GFP and NF-AT-DsRed Reporter Cell Lines Suitable for High-Throughput Scale Detection of Allergic Sensitisation in Array Format and Identification of the ECM–Integrin Interaction as Critical Factor
- First Online:
- Cite this article as:
- Wang, X., Cato, P., Lin, HC. et al. Mol Biotechnol (2014) 56: 136. doi:10.1007/s12033-013-9689-x
- 1k Downloads
We have previously described a microarray platform combining live basophils with protein arrays suitable for high-throughput detection of sensitisation against allergens. During optimisation of this technique, we observed severe losses of adhering cells during the washing steps, particularly after activation. In order to preserve cell binding, we tested the cell adhesion characteristics of different extracellular matrix proteins: human collagen I, fibronectin (FN) from bovine plasma and laminin (LN). FN was more effective than LN and collagen. Cell detachment after activation was in part due to reduced surface expression of VLA-4, the main ligand for FN, which was significantly decreased within 15 min of stimulation with 1 μg/mL calcium ionophore A23187, reaching a minimum after 2 h then slowly recovering. These optimised conditions were used for testing of well-characterised sera from allergic patients using two newly developed rat basophil leukaemia stable reporter cell lines (RBL NF-AT/GFP and RBL NF-AT/DsRed), which both express the human high-affinity IgE receptor alpha chain (FcεRIα). Both cell lines were able to detect sensitisation to specific allergens showing the expected bell-shaped dose–response curve, and correlated (R2 = 0.75) with the standard beta-hexosaminidase assay, which is not suitable for an array format.
KeywordsRBL Basophil array Adhesion Fibronectin VLA-4 Reporter system GFP DsRed
Basophils and mast cells play a central role in allergic reactions . Crosslinking of Immunoglobulin E (IgE) bound to cell surface high-affinity IgE receptor (FcεRI) with antigen triggers intracellular signalling events, leading to multiple cellular responses involving cell degranulation and release of various chemical mediators such as histamine, chemokines, lipid mediators, and cytokines that cause inflammation and common allergic symptoms .
Laboratory diagnosis of allergy is usually performed by measuring levels of allergen-specific IgE (sIgE) in patients’ serum. Traditionally each suspected allergen is tested individually, requiring larger amounts of serum and increasing the overall cost. More recently, technological advances have enabled the simultaneous assessment of specific IgE levels for up to a hundred of allergens, either using multiplex flow cytometry  or allergen arrays [4, 5, 6].
While these new technologies in combination with recombinant or purified allergens allow component resolved diagnosis of allergic sensitisation , there are still unsolved issues regarding the clinical relevance of specific IgE measurements, as the existence of specific IgE binding does not always correlate with clinical symptoms. False-positive results are obtained e.g. due to cross-reactivities with pan-allergens or IgE directed against cross-reactive carbohydrate determinants, which may not have the ability to engage surface-bound IgE productively on mast cells and basophils, and are thus often clinically irrelevant [8, 9].
A recent study found that the concordance between sIgE measurements and skin prick tests (SPT) in allergy to cow’s milk and hen’s egg is unexpectedly low . Similar inconsistencies between tests have also been reported for allergy to hymenoptera venom . As a result, diagnosis of Type 1 Allergy is still best performed using a combination of sIgE, SPT, clinical history, and, in the case of food allergy, oral challenge tests .
Thus there is clearly a need for more clinically relevant diagnostic methods, while taking advantage of component resolved diagnosis and the medium- to high-throughput capacity of novel technologies. In a previous study, we have demonstrated that humanised rat basophil leukaemia (RBL) cell lines could be sensitised by human IgE or sera from allergic patients and degranulation monitored in a cell protein microarray system containing anti-IgE or matching allergen . During the optimisation of the procedure, we observed that in the absence of fixation, sensitised basophils detached from the array surface during washing after activation. The degree of detachment strongly depended on the type of stimulation used, whether proteins were used for coating and the washing protocol.
Although receptor crosslinking is central to the degranulation cascade, other processes potentiate the cell activation process. The adhesion of mast cells and basophils to extracellular matrix (ECM) proteins such as fibronectin (FN), collagen (CO), laminin (LN) or vitronectin (VN) for instance has been shown to play an important role in enhancing degranulation of rat basophil leukaemia RBL-2H3 cells [14, 15, 16]. RBL cells do not adhere to uncoated surfaces or show any adherence in the absence of Ca2+ . RBL adherence to ECM proteins is mediated by specific cell surface receptors that belong to the integrin family .
Many integrins bind the tripeptide Arg-Gly-Asp (RGD) sequence present on ECM proteins, originally described on FN . The β1 integrin subfamily in particular includes receptors that bind to the ECM proteins FN, CO and LN, and have been designed VLA (very late antigen) . VLA-4 and VLA-5 have been shown to mediate RBL-2H3 cell adhesion to FN . VLA-4 recognises the CS-1 sequence present in the type III connecting segment (IIICS) of the cell-binding domain of FN, while VLA-5 recognises the RGD sequences [16, 19].
Our long-term aim is to develop a fully automated basophil array; in order to achieve this, two major obstacles had to be overcome: the cell detachment during the washing steps and the development of reporter systems suitable for an array format. While a NF-AT–luciferase reporter system has been recently developed by Nakamura et al.  for diagnostic purposes, luciferase-based systems require cell lysis before measurement of luciferase activity. However, a lysis step will disconnect the signal from its exact position on the array and is, therefore, not suitable in this format. In this work, we describe optimised conditions for the use of newly generated reporter systems which are suitable for use in high-throughput formats such as allergen arrays, providing a biological readout.
Materials and Methods
Cells and Cell Culture
All experiments were performed with rat basophilic leukaemia cell line (RBL-703/21) transfected with the human FcεRI receptor α chain generated at Paul-Ehrlich-Institut (Langen, Germany; kindly donated by Stefan Vieths and Lothar Vogel). Cells were grown in MR 80/20 medium: 80 % (v/v) Minimal Essential Medium (MEM; GIBCO) plus 20 % (v/v) RPMI-1640 medium (Sigma-Aldrich,UK) supplemented with 5 % (v/v) heat-inactivated Foetal bovine serum (FBS; GIBCO, UK), 2 mM l-glutamine, 50 IU/mL penicillin and 50 mg/mL streptomycin (Invitrogen, Paisley, UK). The transfected cells were cultured in the presence of 1 mg/mL geneticin sulphate (G-418; Sigma-Aldrich) at 37 °C in a humidified atmosphere containing 5 % CO2. Recombinant Allergens were from Greer Labs (Canada).
Binding of Cells to Array Slides
Polyclonal goat anti-human IgE 0.5 mg/mL (Biosource, Camarillo, CA, USA) was hand printed (~0.1 μL/spot) onto 16-Pad FAST slides (Whatman Schleicher & Schuell, Dassel, Germany). The diameter of each spot was approximately 450 μm. The slides were then blocked with 3 % (w/v) BSA in PBS for 30 min at 37 °C, dried, and stored until use at room temperature. Before incubation with the cells, slender strips of adhesive tape (Gene-Frame, ABgene, Epsom, UK) were attached to margins of the coated slide surface to produce a rectangular incubation area for cell/protein interaction. The IgE- or serum-sensitised rat basophilic cell lines were harvested and re-suspended in 0.25 % (w/v) BSA in Dulbecco’s PBS (DPBS) (Lonza, Belgium) at a density of 2.4 × 104 basophils/100 μL and transferred onto the 16-Pad FAST slide. Cells were incubated for 2 h at 37 °C in a humidified atmosphere with 5 % CO2. The slides were gently swirled manually for 10 s every 30 min. After incubation, unbound cells were washed off the slide by loading onto a slide rack and washing in a reservoir containing PBS at room temperature with gentle immersion and lifting for 5 s twice manually, followed by fixation for at least 1 h in PBS containing 1 % (v/v) formaldehyde. Bound cells on FAST slides were observed to be wet using a conventional light microscope with a 4× objective lens. Images were captured and quantified using Image Pro Plus software (Media Cybernetics, Bethesda, MD, USA) equipped with a Nikon Micropublisher 3.3 RTV Labophot-2 camera (Qimaging, Surrey, Canada) and were processed using Adobe Photoshop software.
Surface Expression of Human FcεRI Receptor
In order to optimise culture conditions leading to the surface expression of the human high-affinity IgE receptor, RBL-703/21 cells were incubated with increasing concentrations of IgE for different lengths of time. Normal rat serum was heat-treated (55 °C, 30 min) to inactivate complement and added to cells in medium at 10 % (v/v) in order to reduce potential background staining caused by non-specific antibody binding. Surface expression of FcεRIα was detected by mouse IgG2b CRA-1-FITC monoclonal antibody (CosmoBio, Tokyo, Japan) which is specific for the α-chain of human FcεRI and does not compete for IgE binding. FITC-labelled mouse IgG2b monoclonal antibody specific for keyhole limpet haemocyanin (KLH) was used as the isotype control (R&D Systems, UK). FITC-labelled antibodies (10 μg/mL) were added to 1 × 105 RBL-703/21 in medium and incubated at 4 °C for 30 min. These steps were performed under aseptic conditions in the dark. Samples were then spun at 250 g for 5 min and washed in DPBS (without Ca2+/Mg2+) twice to remove unbound antibody. Finally, cells were fixed in 0.5 % (v/v) formaldehyde in DPBS and analysed in a FACScan flow cytometer (Beckman-Coulter, USA).
FN from bovine plasma (Sigma-Aldrich, UK) was coated at concentrations ranging from 0.625 to 20 μg/mL in a Maxi-Sorp 96-well Elisa plate (Nunc, UK), with one 0.5 % (w/v) bovine serum albumin (Millipore, UK) coated (negative control) and one uncoated well (blank control) at 4 °C overnight prior the assay. 1 × 106 cells/mL of RBL-703/21 were sensitised with human IgE (1 μg/mL) (AbD Serotec, UK) o/n at 37 °C/5 %CO2. After 16 h, untreated and IgE-sensitised cells were re-suspended in serum-free medium to a final concentration of 1 × 106 cells/mL. The pre-coated ECM plate was blocked with 0.5 % (w/v) BSA (Millipore, UK) at room temperature for 1 h. One group of the normal cells and the human IgE-sensitised cells were induced with calcium ionophore A23187 (1 μg/mL) (Sigma-Aldrich, UK) and anti-human IgE (2 μg/mL) (Vector Labs, UK) separately, cell suspension (100 μL) were immediately added to each well of the plate, each condition in triplicates, and incubated for 2 h at 37 °C/5 %CO2. After incubation, the unattached cells were removed by washing by aspiration and replacing with fresh MR80/20 medium three times. The bound cells were stained with 0.1 % (m/v) crystal violet (Millipore, UK) for 10 min, lysed with 10 % (v/v) acetic acid (Fisher, UK) and absorbance was read by a microplate reader (Model 680 XR, Bio-Rad, USA) at 550 nm.
Immunofluorescence Flow Cytometry
The surface expression of VLA-4 integrin was analysed by immunofluorescence flow cytometry. Untreated and IgE-sensitised cells were harvested and resuspended as described above. After washing, cells were incubated in the absence or presence of calcium ionophore A23187 (1 μg/mL) (Sigma-Aldrich, UK) or polyclonal goat anti-human IgE (2 μg/mL) (Vector Labs, UK) at 37 °C/5 %CO2 for 0, 5, 30, 60, 120, 180 and 240 min, then incubated with 2.5 μg/mL fluorescein isothiocyanate (FITC)-conjugated anti-human VLA-4 integrin antibody (Novus, UK) at 37 °C for 1 h. After two washes, cells were fixed with 0.5 % (v/v) formaldehyde (Sigma-Aldrich, UK) on ice, and analysed in a FACScan flow cytometer (Beckman-Coulter, USA). Data were processed and analysed using Weasel software (WEHI, Australia).
Construction of Reporter Cell Lines
The forward primer 5′-ATAATTAAAGATCTGCAGAACTGTGAATGCGCAAACC-3′ and reverse primer 5′-ATAATTAATGAATTCGAGCTCGGTACCCGG-3′ were used to amplify the NF-AT enhancer from pNF-AT-hrGFP plasmid (Stratagene, US), with overhangs and BglII and EcoRI restriction sites, respectively (underlined). The forward primer 5′-ATATAATAGAATTCACCATGGATAGCACTGAGAACGTCATCAAG-3′ and reverse primer 5′-ATATTTAATCTCGAGCTACTGGAACAGGTGGTGGCGG-3′ were used to amplify the DsRed-Express2 gene from pDsRed-Express2-1 plasmid (Clontech, US) with overhangs and EcoRI and XhoI restriction sites, respectively (underlined). Cycling conditions used were: Initial denaturation 94 °C 600 s, followed by 35 cycles of denaturing (94 °C, 60 s), annealing (61 °C, 60 s) and extension (72 °C, 60 s) with a final extension (72 °C, 600 s), on a GeneAMP PCR system 9700 (Applied Biosystems, USA). PCRs were performed using AmpliTaq Gold DNA Polymerase (Invitrogen, UK). To construct the NF-AT/DsRed plasmid, the NF-AT enhancer and DsRed-Express2 PCR products were sequentially ligated into pUB6/V5-His A (Invitrogen, UK) plasmid using T4 DNA ligase (Promega, UK) after digestion of vector with appropriate restriction enzymes as indicated above. DNA sequences were verified by DNA sequencing (Sigma-Aldrich, UK). pNF-AT-hrGFP and NF-AT/DsRed plasmids were amplified in DH5α E. coli and purified using a Plasmid Maxi Kit (QIAGEN, Germany) using standard molecular biology protocols and following the manufacturer’s supplied instructions, respectively. Plasmid purity was assessed by OD260/280 on a Nanodrop 1000. Each plasmid was transfected into RBL-703/21 cells by electroporation. In a 4-mm electroporation cuvette (Bio-Rad, US), 15 μg plasmid was added to 4 × 106 cells in 100 μL MR80/20 medium. Electroporation was carried out at 250 V, 250 μF in a GenePulser Xcell (Bio-Rad, UK). Cells were transferred from the cuvette into 6-well plates containing cell culture medium. 3 days after transfection, stable transfectants were selected using 1 mg/mL hygromycin B (Invitrogen, UK) for pNF-AT-hrGFP transfectants and 20 μg/mL blasticidin S (Invitrogen, UK) for NF-AT/DsRed transfectants. After 6-week selection by hygromycin B, or 1-week selection using blasticidin S, cells were activated using 1 μg/mL human IgE (AbD Serotec, UK) and 2 μg/mL anti-human IgE (Vector Laboratories, UK), and single fluorescent cells were sorted into each well of 96-well plates using a MoFlo cell sorter (Beckman Coulter, USA). The most responsive clones, as indicated by the highest fluorescence after activation, were chosen and maintained in 1 mg/mL hygromycin B or 20 μg/mL blasticidin S.
Reporter gene activation was measured in 384-well plates due to the unavailability of suitable instrumentation with the ability to image array-bound fluorescent cells in a wet environment at sufficiently high resolution (see “Discussion”). The wells were pre-coated with 20 μg/mL FN at 4 °C overnight. Cells were then seeded at a density of 50,000 cells/well and sensitised overnight with 1:20 or 1:40 diluted serum of allergic patients or monoclonal myeloma IgE (1 μg/mL) in 30 μL of MR 80/20 medium. The next day, the medium was replaced with fresh cell culture medium to remove unbound IgE before adding recombinant Bet v1, Phl p1 (both from Biomay AG, Vienna, Austria), Par j2 allergen (Bial Industrial Farmacéutica S.A, Bilbao, Spain) or 2 μg/mL polyclonal anti-human IgE antibody (Vectorlabs, Peterborough, UK) as positive control for induction. Activation of reporter cell lines was measured after 20 h using a Typhoon Trio Scanner (Amersham Bioscience, Sweden) at Ex:Em: 488/520 nm for NF-AT-hrGFP-transfected cells, and Ex/Em: 532/580 nm, for NF-AT-DsRed-transfected cells. Fluorescence intensity in each well was quantified and analysed by Image Quant software (Gel Life Science, USA). Fluorescent images were taken with an EVOS fl microscope (Advanced Microscopy Group, Mill Creek, Washington) at the indicated magnification.
Statistical analysis was performed using a two-tailed Student’s t test. p values <0.05 were considered statistically significant (*p < 0.05; **p < 0.01, or ***p < 0.001). Multiple comparisons were performed with ANOVA and post-hoc Tukey or Bonferroni tests using IBM SPSS Statistics Version 16 (IBM, NY, USA).
Effect of IgE Sensitisation on FcεRIα Chain Expression and Array Binding
Cell Density Experiment
Inclusion of FN Increases Adhesion
VLA-4 is Strongly Downregulated by A23187 but not by IgE-Dependent Activation
The strong, progressive downregulation of VLA-4 after A23187, but not anti-IgE stimulation, may explain the stronger and almost complete detachment of cells seen with ionophore stimulation. To the best of our knowledge, this is the first report of a differential VLA-4 integrin downregulation on RBL cells upon receptor-mediated versus non-receptor-mediated stimulation.
Differential Effects of FN, CO and LA on Cell Attachment After Activation
NF-AT-GFP and NF-AT-DsRed are Induced After Allergen-Dependent IgE Crosslinking
Based on the results shown above, all subsequent experiments were performed in the presence of 20 μg/mL FN. Our next aim was to use these optimised conditions with suitable RBL reporter cell lines, as the array format requires that the activation marker used for assessment of activation remains confined to the cells, in contrast to soluble mediators such as histamine or beta-hexosaminidase, which are released by activated RBL cells.
As shown in Fig. 6a, fluorescence levels showed a dose-dependent bell-shaped curve with a broad optimum. Figure 6b illustrates the related set of control conditions, with unstimulated cells (cell only), cells sensitised with IgE but unstimulated (IgE only), unsensitised and anti-IgE-stimulated cells (anti-IgE only), IgE-sensitised and anti-IgE-stimulated cells (IgE+anti-IgE), serum-sensitised, unstimulated cells (serum only), and serum-sensitised anti-IgE-stimulated cells (serum+anti-IgE). These results show that an increase in GFP fluorescence (Fig. 6c) is only obtained when IgE-sensitised cells are activated by crosslinking agents (either allergen in Fig. 6a or anti-IgE antibodies in Fig. 6b). There was a small, but statistically significant fluorescence increase in serum treated but unstimulated cells, which may be attributed to low-level NF-AT translocation caused by monomeric IgE binding to the high-affinity receptor . However, the fact that we did not find any effect of incubation with monoclonal myeloma IgE seems to exclude this explanation. As we did not find a similar increase with the DsRed reporter system (Fig. 8b), it is more likely to be caused by intrinsic differences between clones, rather than reflecting activation by monomeric IgE.
Overall fluorescence intensity was also increased (Fig 8c). However, the main background due to the used 384-well plastics could not be eliminated or further reduced. In comparison with the use of the GFP reporter, the DsRed reporter showed a sharper, more pronounced bell-shaped activation curve upon stimulation with a serial dilution of the allergen. As can be seen from Fig. 8a, reproducibility between experiments was also good.
Previous work with mast cells and basophils has indicated that FcεRI surface expression is stabilised and increased up to ~17-fold after overnight incubation with its ligand IgE [23, 24]. Significant reduction of surface IgE receptor turnover has also been shown to occur in RBL cells incubated with IgE [25, 26]. However, no previous work has described the kinetics of FcεRI α-chain surface expression in humanised RBL, which is driven by a strong constitutive (e.g. CMV) promoter. Thus our first step in optimising the conditions for the basophil array was to determine the effect of incubation with IgE on surface IgE receptor expression and on the ability to immobilise IgE-sensitised RBL cells on the array surface via an anti-IgE antibody. mRNA levels for human FcεRI α chain were only modestly increased with a ~1.7-fold increase peaking after 4 h (data not shown). This suggests that the increased surface levels of FcεRI α chain observed in Fig. 1, which peaked after 8–10 h, were mainly due to a stabilisation of the receptor at the protein level rather than from induced changes in mRNA expression, in line with the findings of Furuichi et al. .
In many cells, stress fibres emanate from distinct areas of the plasma membrane known as focal adhesions, where clusters of integrin receptors bind to ECM proteins. Focal adhesion proteins at the intracellular surface of the plasma membrane include vinculin, paxillin and talin, which are known to link F-actin with integrins . Adherence to ECM-coated surfaces results in cell spreading, reorganization of cytoskeletal F-actin and the formation of surface ruffles . It was reported that treatment of RBL-2H3 cells with Ca2+ ionophore A23187 or anti-human IgE leads to redistribution of vinculin to the cytoskeletal fraction . Although both anti-human IgE and Ca2+-ionophore can lead to changes of the cell surface morphology, our adhesion experiment results showed that the surface levels of the integrin VLA-4 behave very differently between the receptor- and non-receptor-mediated activation pathways. As a result, while inclusion of FN in the array is important in reducing cell losses during the washing steps, A23187 but not anti-IgE stimulation leads to a pronounced RBL cell detachment after activation even in the presence of FN. Therefore, A23187 is clearly not a suitable positive control for our RBL array. There are several reports describing an increased attachment of mast cells [16, 29, 30] to ECM proteins after activation. In such reports, attachment of primary mast cells to ECM proteins was found to depend on pre-activation, as in the case of bone marrow-derived murine or peritoneal rat mast cells, via PMA [16, 31]. However, such pre-activation is not necessary for RBL [14, 16] or MCP-5 cells  which are able to attach spontaneously. Thompson et al.  assessed attachment of A23187 or antigen-stimulated mast cells to LN and found that approximately 20 % of mast cells attached to LA; however, they did not study the detachment of previously attached cells after stimulation. Thus the apparent discrepancies between reports in the literature describing increased attachment to ECM after activation and our own, which clearly demonstrate pronounced detachment after stimulation with A23187, are due to differences in the protocol.
Another perhaps unexpected result was the ability of CO and LA to bind the RBL cells and prevent losses during the washing step. Ra et al.  have found that while RBL-2H3 bound to VN and FN, they did not bind to CO and LA. However, binding of RBL-2H3 cells to type I CO and LA, although less pronounced than to FN, has been described by other authors [14, 32]. Interestingly, none of the three ECM proteins appeared to increase cell attachment after IgE-dependent activation. Cell loss of CO-attached cells after A23187 treatment was very similar to FN; however, cell loss with LA was far less pronounced. This suggests that the corresponding integrins, which partially overlap between CO and LA but not with FN, are also differentially regulated by A23187 activation in RBL cells.
Regarding the use of the newly developed reporter systems, both GFP and DsRed reporter cell lines showed the expected bell-shaped dose-dependent response. A longer incubation of the reporter cells is needed in comparison with the NF-AT–Luciferase system, which is described as optimal after 3 h . While we were able to detect increases in fluorescence as early as 8-h post stimulation, these changes increased steadily for at least 20 h. We do not know whether this reflects an issue of sensitivity of detection or differences in the rate of synthesis or folding of the different reporter proteins. However, the main problem here was the low signal-to-noise ratio due to the high background, ranging from 2- to 5-fold even under best stimulation conditions. Our final aim is to employ the reporter cell lines using allergen arrays with several hundreds of recombinant allergens or food extracts in parallel as previously used for Immunoglobulin determination . However, the use of conventional laser array scanners is not compatible with the use of cells growing in medium, and commercially available laser microscopes do not have the necessary width of field of view to record the patterns of cellular activation at the array level. Thus our experiments with the fluorescent reporter cell lines had to be performed using 384-well plates and a Typhoon scanner, which can be used with cells immersed in aqueous culture medium. This type of scanner does not have the necessary resolution at the cellular level and the plastic materials used in 384-well cell culture plates have unacceptably high background fluorescence for this application. We are currently developing a microscope with properties which make them more amenable to the measurement of fluorescence in aqueous environment by combining high numerical aperture optics and tiling of mirror scans to obtain a high-resolution image on a large surface.
In summary, we have defined optimal conditions for the use of humanised RBL cells in array format and shown that the inclusion of FN is essential in order to avoid cell losses during the washing steps. A23187 is an unsuitable positive control in this format due to a fast downregulation of VLA-4 and subsequent detachment of cells. Using these optimised conditions, both GFP and DsRed humanised reporter cell lines were able to detect the presence of allergen-specific IgE in the sera of patients and showed the expected bell-shaped dose–response.
Our further efforts will be geared towards implementing these optimised conditions into an automated detection system which can be used for high-throughput assessment of productive IgE–allergen interactions in array format.
This study was partially funded by BBSRC follow-on Grant BB/FOF/268. The humanised RBL cell line clone 703/21 was kindly donated by Stefan Vieths and Lothar Vogel (Paul-Ehrlich-Institut, Langen, Germany). Sera of allergic patients were generously made available by Adriano Mari, Latina, Italy. We also would like to thank Mike Somekh (University of Nottingham, UK) for helpful discussions regarding optical engineering.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.