This is the accepted version of the following article: Hassan A.H.A., Bergua J.F., Morales-Narváez E., Mekoçi A.. Validity of a single antibody-based lateral flow immunoassay depending on graphene oxide for highly sensitive determination of E. coli O157:H7 in minced beef and river water. Food Chemistry, (2019). 297. 124965: - . 10.1016/j.foodchem.2019.124965, which has been published in final form at https://dx.doi.org/10.1016/j.foodchem.2019.124965 © https://dx.doi.org/10.1016/j.foodchem.2019.124965. This manuscript version is made available under the CC-BY-NC-ND 4.0 license http://creativecommons.org/licenses/by-nc-nd/4.0/ 1 Validity of a single antibody-based lateral flow immunoassay depending on graphene oxide 2 for highly sensitive determination of E. coli O157:H7 in minced beef and river water 3 Abdelrahim Hussein Abdelazeem HASSANa,b, José Francisco BERGUAa, Eden MORALES- 4 NARVÁEZc, Arben MEKOÇIa,d,* a 5 Nanobioelectronics & Biosensors Group, Catalan Institute of Nanoscience and Nanotechnology 6 (ICN2), CSIC and BIST, Campus UAB, Bellaterra, 08193 Barcelona, Spain b 7 Department of Food Hygiene and Control, Faculty of Veterinary Medicine, Beni-Suef University, 8 Beni-Suef, 62511, Egypt c 9 Biophotonic Nanosensors Laboratory, Centro de Investigaciones en Óptica A. C, Loma del Bosque 10 115, Lomas del Campestre León, Guanajuato, 37150, Mexico d 11 ICREA, Institució Catalana de Recerca i Estudis Avançats, Pg. Lluís Companys 23, 08010 12 Barcelona, Spain 13 *E-mail: arben.merkoci@icn2.cat 14 15 16 17 18 19 20 21 22 23 24 25 26 1 27 Abstract 28 Considering the health risks of E. coli O157:H7 presence in food and water, an affordable and 29 highly sensitive detection method is crucial. Herein, we report the first use of a single antibody- 30 based fluorescent lateral flow immunoassay (FLFIA) depending on non-radiative energy transfer 31 between graphene oxide and quantum dots for determination of E. coli O157:H7 in beef and river 32 water. FLFIA showed a high sensitivity rate thousand-fold better than the conventional lateral flow 33 (LF). In inoculated minced beef and river water samples, the limits of detection were 178 and 133 34 CFU g-1 or mL-1, respectively. Besides, it presented a high selectivity in the presence of other 35 possible interfering bacteria. The single antibody approach reduced the assay cost to 60% less than 36 the conventional LF. Alongside, the results could be read by portable LF readers or smartphones. 37 These advantages offer FLFIA as a promising technology for pathogen detection in food and water. 38 Keywords 39 Graphene oxide, fluorescent lateral flow, E. coli O157:H7, food safety, minced beef, water quality 40 Chemical compounds studied in this article 41 Phosphate buffered saline (PubChem CID: 24978514); Tween-20 (PubChem CID: 443314); 42 Graphene Oxide (PubChem CID: 124202900); Streptavidin-conjugated CdSe/ZnS quantum dots 43 (Qdot TM 655) (PubChem CID: 121237577) 44 45 46 47 48 49 50 51 52 2 53 1. Introduction 54 Foodborne and waterborne pathogens, mostly bacteria, may get into our bodies through 55 contaminated food and water leading to several health disorders varying from mild diarrhoea to 56 death, and great losses in productivity as well. E. coli O157:H7 is the most frequently reported 57 serotype of Shiga toxins-producing E. coli (STEC) in foodborne-linked hospitalizations and deaths 58 in the United States (Scallan et al., 2011). Beef has been incriminated in most food infection 59 outbreaks by E. coli O157:H7 (CDC, 2009). However other sources such as drinking water, dairy 60 products and vegetables were previously reported as well (Olsen et al., 2002, Islam, Doyle, Phatak, 61 Millner and Jiang, 2004, Lorusso et al., 2011, Tsiraki, Yehia, Elobeid, Osaili, Sakkas and Savvaidis, 62 2018). The health problems that could be induced by E. coli O157:H7 infection range from mild 63 watery diarrhoea to life-threatening conditions such as haemolytic uremic syndrome and 64 haemorrhagic colitis especially in children and the elderly (Jay, 2000). Considering the health risks 65 of E. coli O157:H7 and its impact on food safety, rapid, affordable and highly sensitive methods of 66 detection are necessary to monitor food and water contamination to protect the consumers from the 67 danger of that foodborne hazard. 68 The currently available methods of E. coli O157:H7 detection that depend on culturing and then 69 biochemical and serological examination usually take a couple of days to be completed, while 70 molecular biology-based techniques might be required for confirmation of the results. Nevertheless, 71 such conventional methods are reliable and quite accurate, they are not user-friendly as they require 72 well-trained technicians, and relatively sophisticated laboratory equipment, as well as their high 73 costs (Johnson, Brooke and Fritschel, 1998, Ngwa, Schop, Weir, León-Velarde and Odumeru, 2013, 74 Zhou, Zou, Li, Sun, Ren, and Li, 2018). Immunoassays became one of the most popular approaches 75 in analytical determination of countless kinds of pathogens in various samples, since they are 76 moderately sensitive and selective. Nonetheless, immunoassays such as ELISA and microarrays are 77 laboratory-based techniques that require multiple complex procedures to be done by well-trained 78 operators, as well as they detect E. coli O157:H7 in food at a limit of detection (LOD) ranges from 3 79 105 to 106 colony forming units per mL or g (CFU mL-1 or g-1) or even higher (Firstenberg-Eden 80 and Sullivan, 1997, Arbault, Buecher, Poumerol, and Sorin, 2000, Shen et al., 2014, Zhaohui, 81 Chunyang, Yingchun, and Yanbin, 2017, Kim, Jo, Mun, Noh, and Kim, 2018). Conversely, lateral 82 flow (LF) immunoassays are one of the most important analytical tools nowadays, since they are 83 simple, robust, portable, and rapid devices. Though, those conventional LF immunoassays -which 84 based on gold nanoparticles, latex beads, or quantum dots as labels- always need three antibodies; 85 one for capturing the bacteria (conjugate pad antibody), a second one for detecting the bacteria (test 86 line antibody), and a third one as a control line antibody, which means extra costs spent by such 87 devices (Berg et al., 2015, Zhang et al., 2015, Kim et al., 2018). Moreover, their LOD of E. coli 88 O157:H7 in water and minced beef is about 105 CFU mL-1 or g-1 (Hassan, de la Escosura-Muñiz and 89 Merkoçi, 2015). However, it has been assumed that exposure to < 100 cells of E. coli O157:H7 is 90 enough to induce infection in humans. As the Food and Agriculture Organization of the United 91 Nations and the World Health Organization (FAO/WHO, 2018) reported numerous food poisoning 92 outbreaks by E. coli O157:H7 at doses as low as 5 CFU / g of food. So, they stated that the presence 93 of E. coli O157:H7 at or above one CFU / 25 g constitutes a risky food commodity. Accordingly, 94 the detection of this dangerous pathogen by conventional LF assays might result valueless in 95 particularly demanding contexts. Consequently, the food and environment hygienists are in need to 96 another simple, portable and rapid device that must be affordable, highly sensitive, and highly 97 specific for rapid in-situ determination of E. coli O157:H7 in complex food matrices under the field 98 conditions. 99 Our group has been studying the quenching capabilities of graphene oxide (GO) based on the 100 fluorescence resonance energy transfer (FRET), and its interaction with photoexcited quantum dots 101 (QDs) (Morales-Narváez and Merkoçi 2012, Morales‐Narváez, Hassan and Merkoçi, 2013, 102 Morales-Narváez, Naghdi, Zor, and Merkoçi, 2015; Cheeveewattanagul et al., 2017, Zamora- 103 Gálvez, Morales-Narváez, Romero and Merkoçi, 2018). We had previously patented a highly 104 sensitive pathogen-detection device for the sensing of E. coli in a standard buffer (Morales‐Narváez 4 105 et al., 2013, Patent: EP 13188693.9, 2015). However, using a traditional glass slide-based 106 microarray system as a biosensing platform was quite expensive and not suitable for portability. 107 Therefore, paper-based lateral-flow assay was another low-cost option in another study done by our 108 group (Morales‐Narváez et al., 2015). While, that study was limited to the detection of general E. 109 coli in buffer and bottled water by using QDs/anti-E.coli antibody. Although, assessment of the 110 validity of this GO-based LF immunoassay for detection of pathogenic E. coli O157:H7 in real 111 samples of highly complex matrices such as minced beef and river water is another hot topic worthy 112 to be investigated, since those samples are the main source of human infections by that pathogen. 113 Herein, we report the first exploit of FRET-based quenching properties of GO, and their interaction 114 with QDs for development of a fluorescent lateral flow immunoassay (FLFIA) for determination of 115 the highly pathogenic E. coli O157:H7 in minced beef and river water. The detection part of that 116 strip has two lines; a test line (TL) which composed of CdSe@ZnS QDs/anti-E.coli O157:H7 117 antibody that works as a fluorescent probe and a control line (CL) that composed of only bare QDs. 118 GO is added to the LF strip as a quencher for the fluorescent QDs after adding the sample to divulge 119 the presence of bacteria. If the sample does not have E. coli O157:H7, the test line will be efficiently 120 quenched when adding GO by FRET, since the distance between QDs/Abs (donor) and GO 121 (acceptor) is few nanometres (Gaudreau, Tielrooij, Prawiroatmodjo, Osmond, de Abajo, and 122 Koppens, 2013, Lin et al., 2013). On the other hand, if the sample has E. coli O157:H7, it will be 123 selectively captured by the specific QDs/Abs probe on the test line, then after adding GO, resonance 124 energy transfer is hindered or minimally occurs since the distance between GO and QDs exceeds to 125 more than 20 nm by the bacteria interference (Gaudreau et al., 2013, Lin et al., 2013). 126 Consequently, the fluorescence of QDs on the test line is maintained, and its intensity is 127 correlational to the concentration of the E. coli O157:H7 in the sample. Instead, the control line QDs 128 will be always quenched by GO because this line has not any antibodies to the target pathogen. The 129 principle and reading of FLFIA is fully illustrated in Figure 1. 130 2. Materials and methods 5 131 2.1. Reagents and equipment 132 All commercial reagents were of analytical grade and they were handled according to the safety data 133 sheets provided by the suppliers. Goat polyclonal Anti-Escherichia coli O157:H7 antibody 134 (conjugated with biotin) (LS-C525826-100) was purchased from LifeSpan BioSciences (Seattle, 135 WA, USA), and streptavidin-conjugated CdSe/ZnS quantum dots 655 (QDs) (Cat. No. Q10121MP) 136 were obtained from Life Technologies (Carlsbad, CA, USA). Phosphate buffered saline (PBS) 137 (PubChem CID: 24978514), bovine serum albumin (BSA), and Tween-20 (PubChem CID: 443314) 138 were purchased from Sigma-Aldrich (Madrid, Spain). Graphene oxide (GO) was bought from 139 Angstron Materials (Ohio, U.S.A.). Escherichia coli O157:H7 (CECT 4783, E. coli O157:H7) and 140 Salmonella enterica subsp. enterica serovar Typhimurium LT2 (CECT 722T, S. Typhimurium) 141 strains were obtained from the Colección Española de Cultivos Tipo (CECT, Valencia, Spain). TS- 142 100 Thermo-Shaker (Biosan, Riga, Latvia) was used as a stirrer for modification of QDs with 143 antibodies. Laminated cards (HF000MC100), nitrocellulose membranes (SHF1800425), and 144 cellulose fibre (CFSP001700) that were used for fabricating FLFIA strips were purchased from 145 Millipore (Billerica, MA, USA). An IsoFlow reagent dispensing system (Imagene Technology, 146 Hanover, NH, USA) was used for dispensing the TL and CL onto the nitrocellulose membrane. A 147 Dahle 533 guillotine (Dahle, Peterborough, NH, USA) was used to cut the FLFIA strips into 6 mm 148 width. JP Selecta 2000210 oven from JP selecta (Barcelona, Spain) was used to dry the strips. A 149 portable ESEQuant lateral flow reader with its software LF-Studio Version 3.3.6 from Qiagen 150 GmbH (Stockach, Germany) were used to measure the intensities of the TL and CL of FLFIA strips. 151 As well as, fluorescent images of FLFIA strips were produced using a Typhoon 9410 Variable 152 Mode Imager (GE, Freiburg, Germany). The intensities of the lines of those fluorescent images 153 were measured using ImageJ 1.46r (Wayne Rasband, National Institutes of Health, Bethesda, MD, 154 USA). PBS (10 mM, pH 7.4) with 0.5% (v/v) Tween-20 containing 1% of BSA fraction V (w/v) 155 was employed as a standard buffer for preparation of bacterial inocula. While, PBS (10 mM, pH 156 7.4) with 0.05% (v/v) Tween-20 was used as a washing buffer. All aqueous solutions were freshly 6 157 prepared in Milli-Q water produced using a Milli-Q system (>18.2 MΩcm−1) purchased from 158 Millipore (Billerica, MA, USA). Scanning Electron Microscopy (SEM) images were obtained by a 159 Magellan 400L High-Resolution SEM (FEI, Hillsboro, OR, USA). 160 2.2. Preparation of minced beef extract and bacterial inocula 161 Minced beef was purchased from a local retail market in Barcelona and analysed by the standard 162 culturing method for the presence of E. coli O157 (ISO 16654:2001). Only negative samples of beef 163 and water were selected to be inoculated with bacteria. Twenty-five g of E. coli O157-free minced 164 beef were homogenized in a sterile stomacher bag with 225 mL of sterile PBS (10 mM, pH 7.4) 165 using a stomacher (Lab Blender 400, Seward, UK) for 3 min. Then the filtrate was used as a diluent 166 for preparation of bacterial suspensions. 167 For preparation of bacterial inocula, freeze-dried cultures of E. coli O157:H7 and Salmonella 168 Typhimurium were revived in a sterile tryptone soy broth (TSB, Oxoid Ltd., UK) and incubated at 169 37 ºC for about 24 h, then transferred onto sterile tryptone soy agar (TSA, Oxoid Ltd., UK) plates 170 for another 24 h at 37 ºC. Stock cultures of both strains were kept on TSA slope tubes for future use. 171 Bacterial cell suspensions were prepared directly from bacterial colonies of TSA plates, during the 172 logarithmic phase, in sterile standard buffer and river water to obtain a bacterial load of 1.5 × 108 173 CFU mL-1 according to McFarland standards (McFarland, 1907) using Densimat densitometer 174 (Biomerieux, Brazil). Afterwards, ten-fold decimal bacterial dilutions (10 to 108 CFU mL-1) were 175 prepared from the original one. Finally, heat killing of the bacteria was done by putting the bacterial 176 suspension in tightly sealed tubes to be placed in a water bath at 80 ºC for 15 min to stop bacterial 177 replication. Regarding minced beef, a suitable volume of heat-killed bacterial suspension in a sterile 178 standard buffer (1.5 × 108 CFU mL-1) was used to prepare ten-fold decimal dilutions of E. coli 179 O157:H7 in minced beef homogenate (10 to 108 CFU g-1). The prepared bacterial dilutions were 180 stored at 4 °C until being used for the assay within two weeks in case of standard buffer and river 181 water. Whereas inoculated minced beef was used without delay to avoid sample deterioration. 182 2.3. Fabrication of FLFIA 7 183 The proposed lateral flow strips were prepared as follows: (a) assembling of the nitrocellulose 184 membrane on the laminated card. (b) Dispensing the QDs/anti-E. coli O157:H7 as a TL and bare 185 QDs as a CL using an IsoFlow reagent dispensing system on the nitrocellulose membrane. For TL, 186 we used a conjugate composed of 4 nM streptavidin-quantum dots 655 and 300 µg/mL biotinylated 187 anti- E. coli O157:H7 polyclonal antibody in standard buffer. The conjugate was prepared through 188 mixing them at 650 rpm/4 °C /30 min. Whereas for CL, we used only 4 nM of streptavidin-quantum 189 dots 655. After line dispensing, the detection pad was kept overnight inside a tightly closed 190 container in the fridge at 4 °C temperature. In the second day, the nitrocellulose membrane (2.5 × 20 191 cm) was homogenously treated with 5 mL of standard buffer, then kept in the fridge for 15 min 192 before drying in the oven at 37 °C for about 3 h. (c) Some pieces of cellulose sample and absorbent 193 pads (≈20 cm each) were saturated sequentially with Milli-Q water, and standard buffer, then they 194 were kept in the oven at 37 °C for overnight until complete dryness. (d) Afterwards, assembling the 195 sample and absorbent pads on the same laminated card. (e) Ultimately, cutting the assembled card 196 using a clean guillotine into strips of 6 mm in width. The strips were kept in a tightly closed plastic 197 container with some drying pearls in the fridge until use for bacteria determination. 198 2.4. Using FLFIA for E. coli O157:H7 detection in standard buffer 199 In order to use the prepared FLFIA strips for detection and quantification of E. coli O157:H7 in 200 various samples, the initial photoluminescence intensities (I1) of both TL and CL were measured 201 using a portable lateral flow reader (Figure 1C). Then 100 µL of previously prepared E. coli 202 O157:H7 suspension of various concentrations in standard buffer was added onto the sample pad of 203 the fabricated strip, the strips were left for about 15 min until complete flow of the sample to the 204 absorbent pad. Afterwards, 100 µL of PBS with 0.05 % tween 20 (v/v) was dispensed on the sample 205 pad as a washing buffer, to remove any kind of intervention. Then, they were left at room 206 temperature for around 10 min until complete flow of the washing buffer. Subsequently, 100 µL of 207 aqueous solution of graphene oxide (GO) 150 µg mL-1 contains 0.1 % Tween-20 (v/v) was 208 dispensed on the sample pad, for revealing the presence of bacteria. A final step of dryness was 8 209 done before reading the final photoluminescence intensities (I2) of both lines using the LF reader. 210 The ratio of the final intensity to the initial one (I2/I1) of the test line (RTL) was used as an estimation 211 for the concentration of the target bacteria in the sample. 212 2.5. Validation of FLFIA in real samples 213 To evaluate the overall performance of the proposed assay in real samples, artificially inoculated 214 minced beef and river water with serial concentrations of E. coli O157:H7 (0, 50, 102, 103, 104, 105, 215 and 106) CFU g-1/mL-1 were used. The same abovementioned procedure used with standard buffer 216 was conducted with real samples as well. Calibration curves were created for each sample type at 217 decimal concentrations of bacteria. The linear regression and coefficient of determination (R2) were 218 calculated for both minced beef and river water. 219 Furthermore, spike and recovery experiment was conducted to distinguish how much the complex 220 matrix of real sample could affect the performance of our FLFIA. Two concentrations of E. coli 221 O157:H7 (103 and 104 CFU mL-1 or g-1) were spiked in each of standard buffer, minced beef, and 222 river water. At least 3 replicates were used in each concentration. The average RTL of spiked minced 223 beef and river water was compared to that of standard buffer at the same concentration to estimate 224 the recovery percentage according to the following equation; recovery %= RTL of real sample/RTL of 225 standard buffer. 226 As well as, the specificity of FLFIA against non-specific pathogen was tested. Salmonella enterica 227 subsp. enterica serovar Typhimurium (S. Typhimurium), a Gram-negative pathogen from the same 228 Enterobacteriaceae family of E. coli O157:H7, was used to conduct the specificity test. In this 229 experiment, we evaluated the response of FLFIA to the presence of S. Typhimurium either alone or 230 in a mixture with E. coli O157:H7, as well as it was compared with blank buffer. Blank (0 CFU mL- 1 231 ), single S. Typhimurium (104 CFU mL-1), single E. coli O157:H7 (102 and 104 CFU mL-1), and 232 two mixtures of both bacterial species (E. coli O157:H7 102 + S. Typhimurium 104 and E. coli 233 O157:H7 104 + S. Typhimurium 102 CFU mL-1) were prepared in standard buffer to conduct such 234 experiment. 9 235 Additionally, the reproducibility of the assay was evaluated by estimating the variation coefficient 236 through calculating the relative standard deviation (RSD %) along different batches of FLFIA strips 237 used throughout the study. 238 3. Results and discussion 239 3.1. Optimization of fluorescence and quenching process 240 Since the currently available conventional LF immunoassays based on gold nanoparticles or latex 241 beads used for E. coli O157:H7 detection in various food and water samples are of high costs (≈ 242 0.30 USD / test strip) and low sensitivity (≈ 105 CFU mL-1 or g-1) (Karakus and Salih, 2013; Hassan, 243 et al. 2015; Luo et al., 2017; Han et al., 2018), the food and water monitoring may require another 244 simple, portable, affordable and highly sensitive device. Herein, we designed a novel fluorescent 245 lateral flow immunoassay based on the interaction between photoexcited molecules and quencher. 246 We exploited streptavidin functionalized CdSe@ZnS QDs, of an approximate diameter of 14±2 nm 247 and a maximum emission wavelength at ≈ 665 nm, as donors of non-radiative energy that makes 248 them a powerful fluorescence agent. As well as, we used GO sheets in the form of water-based 249 dispersion of an average lateral dimension range of 500 nm, an average thickness of approximately 250 1.1 nm and C/O ratio of about one unit (according to manufacturer`s data), as acceptors for the non- 251 radiative energy leading to highly effectual quenching of fluorescence (Morales‐Narváez et al., 252 2013). SEM images illustrated in Figures 2C and 2C1 show GO sheets in water suspension 253 surrounding to E. coli O157:H7 cells. In addition, Figure 2C2 shows bare GO suspension of the 254 same concentration. Since the distance between the QDs and GO is very crucial for non-radiative 255 energy transfer between them as Lin et al. (2013) recorded that quenching is not strongly observable 256 at distances greater than 20 nm, so here the target bacteria (≈ 0.5×2 µm size) acts as a spacer 257 between the donor and the acceptor hindering the photons transfer and keeping the fluorescence of 258 QDs. Figure 2D shows a SEM micrograph of QDs-Ab conjugates are capturing to bacterial cells of 259 E coli O157:H7. 10 260 To get the most suitable photoluminescence, different concentrations of QDs (1.5, 3, 4, 6, 8, 9, and 261 10 nM) were dispensed on nitrocellulose membranes and their intensities were measured by a 262 portable LF reader (data not shown). The LF reader used in this study has an excitation wavelength 263 of 365 nm, and an emission filter of about 670 nm. Hence, 4 nM was chosen as the appropriate 264 concentration that gives about 80% of the dynamic range of the reader (Figure 1C). Additionally, 265 since the concentration of the acceptor molecules should affect the rate of photons transfer from the 266 donor to the acceptor, so, different concentrations of GO suspension in Milli-Q water (60, 70, 80, 267 100, 150 and 200 µg mL-1) with two concentrations of Tween-20 (0.05 and 0.1 %) were 268 investigated to optimize the most suitable quenching conditions. The presence of Tween-20 in the 269 GO suspension aids the process of GO flow through the nitrocellulose. A hundred µL of each 270 concentration was added onto a blank strip, the TL and CL intensities were measured before and 271 after addition of GO. The degree of quenching (I2/I1) was calculated by dividing the final intensity 272 (I2) by the initial one (I1). GO 150 and 200 µg mL-1 with 0.1 % Tween-20 (v/v) achieved the highest 273 quenching rates (I2/I1 ≈0.3-0.4) (Figure 3A). However, GO 150 µg mL-1 with 0.1 % Tween-20 (v/v) 274 was preferred because it achieved the most reliable results afterwards, in terms of steady 275 performance and error rate. In conclusion, 4 nM QDs and GO 150 μg mL-1 with Tween-20 (0.1 % 276 v/v) were the most appropriate condition for proper photoluminescence and quenching of blank 277 strips. 278 Since bacterial cells are much bigger (≈ 0.5 ×2.0 µm, Figure 2A) than other analytes like proteins, a 279 nitrocellulose membrane with big pore size (the diameter of the largest pore in the filtration 280 direction) was essential for our proposed assay. Moreover, there is an inverse relationship between 281 the flow rate and sensitivity of the assay, that means slow flow rate should give highly sensitive 282 assays, because it allows longer time of interaction between the antibody and the target analyte, 283 while fast flow rate reduces the sensitivity. Thus, to develop a highly sensitive assay for a big 284 analyte like bacteria, Hi-Flow 180 nitrocellulose membrane (SHF1800425) of slow flow rate (≈ 180 285 seconds/4 cm) was chosen out of others to develop our assay. Figure 2B demonstrates SEM image 11 286 of HF 180 nitrocellulose membrane used in this study, it proves that the pore size (≈ 8 - 20 µm) is 287 big enough to allow the proper flow of bacteria. Besides, it shows the difficulty of distinguishing 288 between bacterial cells and nitrocellulose tissue by SEM. So, all SEM images of bacterial cells 289 (Figures 2A, 2C and 2D) were prepared on silicon discs not on nitrocellulose. The total cost of each 290 strip of this fluorescent lateral flow assay was previously estimated to be ≈ 0.12 USD (Zamora- 291 Gálvez et al., 2018), which is considered about 60% less than that of conventional lateral flow strip. 292 3.2. Optimization in standard buffer 293 To evaluate the overall performance of the proposed FLFIA, serial concentrations of E. coli 294 O157:H7 (0, 102, 103, 104, 105 and 106 CFU mL-1) in standard buffer were investigated. A hundred 295 µL of each concentration was loaded onto the sample pad of FLFIA strips, then followed by 100 µL 296 of GO 150 µg mL-1 with Tween 20 (0.1 % v/v). A drying step of the strips for almost an hour in an 297 oven at 35 °C before reading them using a portable lateral flow reader was essential because QDs 298 have better photoluminescence capabilities in the solid phase than the liquid one (Shi et al., 2010). 299 Afterwards, the ratio of the final intensity of TL (I2, after addition of GO) to the initial one (I1, 300 before addition of the sample) was calculated and used as an indicator to the presence or absence of 301 E. coli O157:H7. We refer to it in this paper as RTL= I2/I1 of TL. As high RTL (close to one) 302 indicates low quenching rate and high concentration of bacteria, whereas low RTL (close to zero) 303 indicates high quenching rate and low concentration or absence of bacteria. On the other hand, I2/I1 304 of CL = RCL should be unchangeable with varying bacteria concentrations, since there are not any 305 antibodies on the CL. However, CL is essential to prove the successful flow of GO along the strip. 306 The obtained results showed an elevation in RTL with increasing the concentration of bacteria in the 307 standard buffer, which means that the target E. coli O157:H7 is captured by the specific antibody of 308 TL (anti-E. coli O157:H7). However, a similar phenomenon was observed in RCL as well. That 309 indicates some bacterial cells halt over CL and act as a spacer between GO and QDs there, thus 310 leading to non-specific response of CL (Figure 3B). Therefore, a washing step with 100 µL of PBS 311 with 0.05 % Tween 20 (v/v) by dispensing onto the sample pad after complete flow of the bacteria- 12 312 containing buffer along the strip (approximately after 15 min) was suggested to remove any kind of 313 non-specific response before addition of GO. 314 Obviously, this washing step has improved greatly that issue of non-specific response of CL, 315 leading to almost constant RCL with varying concentrations of E. coli O157:H7 in the buffer (Figure 316 3C), while RTL increased progressively with increasing the concentration of bacteria (from zero to 317 105 CFU mL-1) and in a logarithmic manner from 50 to 105 CFU mL -1 with R2 equals 0.9874. Then 318 this response slightly decreased in concentrations higher than 105 CFU mL -1 (Figure 4A, 4B). This 319 decline behaviour in response could be attributed to blocking the nitrocellulose membrane by the 320 enormous number of bacteria that lead to hindering the bacterial flow. A similar phenomenon was 321 previously reported by some literatures such as Hassan et al. (2015) who reported a decline behavior 322 in commercial gold nanoparticles-based lateral flow kits for E. coli O157:H7. 323 To estimate the sensitivity of FLFIA for detection of E. coli O157:H7 in standard buffer, the mean 324 RTL of blank samples plus 3 times its standard deviation (SD) was calculated and used to determine 325 the limit of detection (LOD) of the assay. Fascinatingly, the estimated LOD of FLFIA was 326 calculated to be 57 CFU mL-1 of E. coli O157:H7 in standard buffer (Figure 4A). This achieved 327 LOD by our assay was about thousand-fold better than the conventional lateral flow assays (LOD ≈ 328 105 CFU mL-1 or g-1) that depend on a sandwich-type immunoassay on the TL and a third Ab on the 329 CL (Karakus and Salih, 2013; Hassan, et al. 2015; Luo et al., 2017; Han et al., 2018). However, the 330 proposed FLFIA requires only one antibody on the TL and without any antibodies on the CL. 331 3.3. Specificity of FLFIA 332 The specificity of immunoassays is another crucial parameter of any innovative approach. The 333 obtained results summarized in Figure 4C showed that the average RTL produced by S. 334 Typhimurium (104 CFU mL-1) was lower than the blank`s one. Equally, the mixture of E. coli 335 O157:H7 102 + S. Typhimurium 104 CFU mL-1 gave a response similar to E. coli O157:H7 102 CFU 336 mL-1. Likewise, E. coli O157:H7 104 + S. Typhimurium 102 CFU mL-1 and E. coli O157:H7 104 337 CFU mL-1. That experiment proved the high selectivity of the proposed FLFIA to the target 13 338 pathogen (E. coli O157:H7), without any interferences from non-specific bacteria present in the 339 same medium. 340 3.4. Using FLFIA for determination of E. coli O157:H7 in real samples 341 Even though, investigation of the performance in standard buffer is quite important for the 342 optimization process, the evaluation in complex matrices is vital for validation of new methods. The 343 data illustrated in Figure 5 summarize the performance in real samples. The obtained results in 344 minced beef and river water showed a similar scenario to that of standard buffer. As RTL elevated 345 regularly in a logarithmic response in concentrations from 50 to 105 CFU g-1/mL-1, with R2 equals 346 0.9592 and 0.9542 in minced beef and river water samples, respectively. Then the response slightly 347 declined in concentrations higher than 105 CFU g-1/mL-1 in both sample types, however, it is still 348 within the positive range. The same decline behaviour in response to high concentrations happened 349 with standard buffer, which confirms that this behaviour is due to the blockage of flow by the vast 350 bacterial number in higher concentrations. 351 LOD of FLFIA in minced beef and river water was estimated by calculating the averages RTL of 352 FLFIA strips tested with at least 3 replicates of blank minced beef and blank river water plus 3 353 times their SD. The obtained LOD in minced beef samples was ca. 178 CFU g-1, while it was 354 ca.133 CFU mL-1 in river water ones (Figure 5A). The reduced sensitivity in minced beef and river 355 water than standard buffer is attributed to the matrix effect of real samples. Similar effect of the 356 sample matrix on immunoassays were previously reported by Aydin, Herzig, Jeong, Dunigan, Shah, 357 Ahn (2014), Hassan et al. (2015), Luo et al., (2017), Han et al. (2018) and Kim et al. (2018). 358 However, our achieved LODs in real samples do not affect the reliability of FLFIA and confirm its 359 high sensitivity in comparing with conventional immunoassays. The ability to detect E. coli 360 O157:H7 at such low concentrations without broth enrichment designates that FLFIA could be used 361 to determine as low as one CFU g-1 or mL-1 of E. coli O157:H7 in minced beef and water samples 362 after about 3 h of broth enrichment of the sample, since E. coli O157: H7 could duplicate by mitotic 363 division every 15-20 min under favourable conditions (Buchanan and Klawitter, 1993). 14 364 By comparing FLFIA in terms of LOD with other reported rapid methods, which were depending 365 on sandwich antibody formats, more complicated techniques and/or more expensive approaches for 366 determination of E. coli O157:H7 in various food samples, we noticed the high sensitivity of our 367 costless approach over those more complicated and expensive technologies. For instance, Aydin et 368 al. (2014) reported 250 CFU g-1 as a LOD of E. coli O157:H7 in ground beef using magnetic bead– 369 based immunoassay coupled with tyramide signal amplification after 3 h of enrichment. Hassan et 370 al. (2015) reported E. coli O157:H7 LODs of 457 and 309 CFU g-1 or mL-1 in minced beef and tap 371 water samples, respectively, through using gold nanoparticles-labelled antibody sandwich-based 372 electrochemical detection. Additionally, Song, Li, Liu, Liu (2016) reported an E. coli O157:H7 373 LOD of 105 CFU g-1 or mL-1 in bread, milk and jelly samples using Fluorescein isothiocyanate- 374 based immunosensor. As well as, Luo et al. (2017) compared different immunochromatographic 375 labels for lateral flow assays for E. coli O157:H7 determination in milk. In that study, they reported 376 LODs accounted for 1 × 105, 2.5 × 104, 1 × 103, 5 × 102 CFU mL-1 using gold nanoparticles, 377 quantum dots, fluorescent nanoparticles, and europium chelate nanoparticles as labels, respectively. 378 Eventually, Han et al. (2018) mentioned that the sensitivity of the nanozyme-based LFA depending 379 on a sandwich antibody format developed by them for E. coli O157:H7 was 900 CFU mL-1 in milk. 380 3.5. Spike and recovery test in real samples 381 The results of spike and recovery experiment are summarized in Table (1). The recovery 382 percentages from minced beef ranged from 92.86 to 95.02 %, while those of river water ranged 383 from 95.11 to 97.98 %. Obviously, the extreme complex matrix of beef affects the assay 384 performance more than that of river water. However, it still performs in an admirable way, adequate 385 for real application requirements. Accordingly, these recovery rates demonstrate that this novel 386 approach is a promising device for determination of E. coli O157:H7 in food and water without any 387 interferences from the complex food and water matrices nor other competing microorganisms. 388 3.6. Reproducibility 15 389 Another important parameter for evaluating new analytical technologies is the reproducibility. In 390 this study, for executing all experiments mentioned above, we used different fabrication batches of 391 FLFIA. Among the working range of bacterial concentrations, the FLFIA strips exhibited variation 392 coefficients below 16 % in minced beef and river water. This meets the validation criterion of 393 reproducibility of new immunoassays that was stated by Findlay et al. (2000), who recommended a 394 RSD below 20 % for acceptance of new procedure in terms of reproducibility. 395 3.7. Possibility of smartphone integration 396 To prove the possibility of integration of proposed FLFIA into smartphones without the need to a 397 portable lateral flow reader, another device for reading the line intensities was tried. A fluorescence 398 image Typhoon scanner was used to take pictures of the strips. Then those scanned pictures were 399 analysed using ImageJ 1.46r software to determine line intensities (Figure 1B). A similar procedure 400 with smartphones could be used through a 3D-printed cassette containing an excitation LED for 401 holding the FLIFA strip to enable smartphone camera to capture the fluorescence and then an 402 ImageJ application (smartphone version) be used for analyzing the picture. This proof of concept 403 makes it a highly promising device for automation, portability, and field applications without the 404 need for a highly equipped laboratory. 405 3.8. FLFIA versus traditional methods for detecting E. coli O157 in random food samples 406 Herein, we summarize the whole procedure of using FLFIA to analyze a random unknown food 407 sample for the presence of E. coli O157:H7, in comparing with the traditional method in terms of 408 procedure and assay time. In case of FLFIA, firstly, 25 g or mL of the food sample is homogenized 409 or mixed with 225 mL of a pre-warmed modified tryptone soya broth plus novobiocin (mTSB+N) at 410 41.5 °C ± 1°C and then incubated for 3 h. Meanwhile, the TL initial intensity of FLFIA strip being 411 recorded using a portable LF reader. Then a 100 µL of the incubated sample broth is added onto the 412 sample pad. Wait for 10 minutes to allow sample flow. Subsequently, a 100 µL of washing buffer, 413 followed by a 100 µL of GO solution are added. After strip dryness, record the TL final intensity. If 414 the RTL is < 0.4 indicates a negative sample, while if it is ≥ 0.4 indicates a positive sample. Though, 16 415 CL should be quenched in both positive and negative samples to confirm the successful flow of 416 solutions through the strip pads. Accordingly, the total assay time of FLFIA is only 5 h, including 3 417 h of sample enrichment. 418 On the other hand, in order to detect E. coli O157 in food samples, using the traditional horizontal 419 method stated by the International Organization for Standardization (ISO 16654:2001) more than 60 420 h of sample examination were required to confirm the presence of this pathogen. The detection of E. 421 coli O157 by ISO`s method necessitates four successive stages: a) enrichment, b) separation and 422 concentration, c) isolation and d) confirmation. Briefly, the sample was enriched in nine times the 423 weight in mTSB+N for 6 h and subsequently for a further 12 - 18 h. Then E. coli O157 were 424 separated and concentrated using immunomagnetic beads coated with anti-E. coli O157 antibodies 425 after 6 h and again, if necessary, after a further 12 - 18 h incubation. Afterwards, E. coli O157 426 captured with immunomagnetic particles were subcultured onto cefixime tellurite sorbitol 427 MacConkey agar (CT-SMAC) and the agar plates were incubated at 37 °C /18 - 24 h. Subsequently, 428 typical E. coli O157 colonies (sorbitol negative) were streaked onto nutrient agar (NA) and 429 incubated at 37 °C /18 - 24 h. Eventually, E. coli O157 on NA was confirmed by indole production 430 and agglutination with E. coli O157 antiserum. Thus, the traditional method is laborious, time 431 consuming and of high cost, as well as, it requires well-trained operators and highly equipped 432 facilities. That confirms the advantages of FLFIA over standard traditional methods. 433 4. Conclusions 434 In conclusion, we exclusively developed a fluorescent lateral flow immunoassay based on quantum 435 dots as donors of non-radiative energy and graphene oxide as an acceptor for such energy. We used 436 only a single antibody on the test line to capture the target pathogen, which reduced the total assay 437 cost per strip to be 60% less than the conventional LF. This study is the first report of using that 438 principle for E. coli O157:H7 detection in minced beef and river water. FLFIA achieved outstanding 439 LODs of E. coli O157:H7 (≈133 and 178 CFU mL-1 or g-1 in river water and mined beef, 440 respectively). Theoretically, this indicates the possibility of detecting as low as one CFU mL-1 or g-1 17 441 of E. coli O157:H7 after about 3 h of food sample enrichment in a suitable broth. The detection of 442 E. coli O157:H7 by FLFIA in beef complex matrix designates the ability of their using for other 443 food commodities, as well as for other similar bacterial species with changing the antibody. A 444 portable lateral flow reader was used for reading and quantifying the results. Alongside, analysing 445 the images with ImageJ software was proved to be an alternative way for reading the results with 446 smartphones. FLFIA showed numerous advantages in comparing with the standard traditional 447 method of E. coli O157 detection, as well as against other previously reported rapid methods. 448 Declaration of Competing Interest 449 The authors declare that there is no any conflict of interest in this work. 450 Acknowledgement 451 A.H.A. Hassan gratefully acknowledges the financial support from the Science and Technology 452 Development Fund (STDF), Egypt (Grant 25347). E. M.-N. acknowledges the financial support 453 from CONACYT (Mexico, Grant 293523) and National System of Researchers, CONACYT 454 (Mexico, Grant 74314). ICN2 is supported by the Severo Ochoa program from Spanish MINECO 455 (Grant No. SEV-2017-0706 and MAT2017) and by CERCA Programme/ Generalitat de Catalunya. 456 References 457 Arbault, P., Buecher, V., Poumerol, S., & Sorin, M. L. (2000). Study of an ELISA method for the 458 detection of E. coli O157 in food. In Progress in Biotechnology (Vol. 17, pp. 359-368). Elsevier. 459 https://doi.org/10.1016/S0921-0423(00)80093-3. 460 Aydin, M., Herzig, G. P., Jeong, K. C., Dunigan, S., Shah, P., & Ahn, S. (2014). Rapid and 461 Sensitive Detection of Escherichia coli O157: H7 in Milk and Ground Beef Using Magnetic Bead– 462 Based Immunoassay Coupled with Tyramide Signal Amplification. Journal of Food 463 Protection, 77(1), 100-105. https://doi:10.4315/0362-028X.JFP-13-274. 464 Berg, B., Cortazar, B., Tseng, D., Ozkan, H., Feng, S., Wei, Q., Chan, R.Y.L., Burbano, J., 465 Farooqui, Q., Lewinski, M., & Di Carlo, D. (2015). Cellphone-based hand-held microplate reader 18 466 for point-of-care testing of enzyme-linked immunosorbent assays. ACS nano, 9, 7857-7866. 467 https://doi:10.1021/acsnano.5b03203. 468 Buchanan R.L., & Klawitter L.A. (1992). The Effect of Incubation Temperature, Initial pH, and 469 Sodium Chloride on the Growth Kinetics of Escherichia coli O157:H7: Food Microbiology, 9, 185- 470 196. https://doi.org/10.1016/0740-0020(92)80046-7 471 CDC, Centers for Disease Control and Prevention (2009). Reports of E. coli Outbreak 472 Investigations from 2006-2009, CDC, USA. URL https://www.cdc.gov/ecoli/2006-2009- 473 outbreaks.html. Accessed 03.06.2019. 474 Cheeveewattanagul, N., Morales‐Narváez, E., Hassan, A.-R. H., Bergua, J. F., Surareungchai, W., 475 Somasundrum, M., & Merkoçi, A. (2017). Straightforward Immunosensing Platform Based on 476 Graphene Oxide‐Decorated Nanopaper: A Highly Sensitive and Fast Biosensing 477 Approach. Advanced Functional Materials, 27(38), 1702741. 478 https://doi.org/10.1002/adfm.201702741. 479 Findlay, J. W. A., Smith, W. C., Lee, J. W., Nordblom, G. D., Das, I., DeSilva, B. S., Khan, M. N., 480 & Bowsher, R. R. (2000). Validation of immunoassays for bioanalysis: a pharmaceutical industry 481 perspective. Journal of pharmaceutical and biomedical analysis, 21(6), 1249-1273. 482 https://doi.org/10.1016/S0731-7085(99)00244-7 483 Food and Agriculture Organization of the United Nations/ World Health Organization (FAO/WHO, 484 2008). Shiga toxin-producing Escherichia coli (STEC) and food: attribution, characterization, and 485 monitoring (Report). Microbiological risk assessment series 31, Rome. Available on: 486 www.fao.org/publications. Accessed 03.06.2019. 487 Gaudreau, L., Tielrooij, K. J., Prawiroatmodjo, G. E. D. K., Osmond, J., de Abajo, F. G., & 488 Koppens, F. H. L. (2013). Universal distance-scaling of nonradiative energy transfer to 489 graphene. Nano letters, 13 (5), 2030-2035. https://doi:10.1021/nl400176b 19 490 Han, J., Zhang, L., Hu, L., Xing, K., Lu, X., Huang, Y., Zhang, J., Lai, W. & Chen, T., (2018). 491 Nanozyme-based lateral flow assay for the sensitive detection of Escherichia coli O157: H7 in 492 milk. Journal of Dairy Science. 101:5570-5779. https://doi:10.3168/jds.2018-14429 493 Hassan, A.-R. H. A.-A., de la Escosura-Muñiz, A., & Merkoçi, A. (2015). Highly sensitive and 494 rapid determination of Escherichia coli O157: H7 in minced beef and water using electrocatalytic 495 gold nanoparticle tags. Biosensors and Bioelectronics, 67, 511-515. 496 https://doi:10.1016/j.bios.2014.09.019 497 Islam, M., Doyle, M. P., Phatak, S. C., Millner, P., & Jiang, X., (2004). Persistence of 498 enterohemorrhagic Escherichia coli 0157:H7 in soil and on leaf lettuce and parsley grown in fields 499 treated with contaminated manure composts or irrigation water. Journal of Food Protection, 67, 500 1365-1370. 501 ISO 16654:2001. Microbiology of food and animal feeding stuffs - Horizontal method for the 502 detection of Escherichia coli O157. URL https://www.iso.org/standard/29821.html. Accessed 503 03.06.2019. 504 Jay, J. (2000). Microorganisms in foods. In Modern Food Microbiology (6 ed.). New York: Van 505 Nostrand Reinhold. 506 Johnson, J., Brooke, C., & Fritschel, S. (1998). Comparison of the BAX for screening/E. coli O157: 507 H7 method with conventional methods for detection of extremely low levels of Escherichia coli 508 O157: H7 in ground beef. Applied Environmental Microbiology, 64, 4390-4395. 509 Karakus, C., & Salih, B. A. J. (2013). Comparison of lateral flow immunoassays (LFIA) for the 510 diagnosis of Helicobacter pylori infection. Journal of Immunological Methods, 396, 8-14. 511 https://doi.org/10.1016/j.jim.2013.08.010. 512 Kim, S. U., Jo, E.-J., Mun, H., Noh, Y., & Kim, M.-G. (2018). Ultrasensitive Detection of 513 Escherichia coli O157:H7 by Immunomagnetic Separation and Selective Filtration with Nitroblue 514 Tetrazolium/5-Bromo-4-chloro-3-indolyl Phosphate Signal Amplification. Journal of Agricultural 515 and Food Chemistry, 66, 4941–4947. https://doi:10.1021/acs.jafc.8b00973 20 516 Lin T. N., Huang L. T., Shu G.W., Yuan C. T., Shen J. L., Lin C. A. J., Chang W. H., Chiu C. H., 517 Lin D. W., Lin C. C., & Kuo H. C. (2013). Distance dependence of energy transfer from InGaN 518 quantum wells to graphene oxide. Optics letters, 38 (15), 2897-2899. 519 https://doi.org/10.1364/OL.38.002897 520 Lorusso, V., Dambrosio, A., Quaglia, N. C., Parisi, A., Lasalandra, G., Mula, G., Virgilio, S., 521 Lucifora, G., Dario, M., & Normanno, G. (2011). Development of a multiplex PCR for rapid 522 detection of verocytotoxin-producing Escherichia coli O26 in raw milk and ground beef. Journal of 523 Food Protection, 74, 13-17. https://doi:10.4315/0362-028X.JFP-10-201. 524 Luo, K., Hu, L. Guo, Q. Wu, C. Wu, S. Liu, D. Xiong, Y., & Lai, W. (2017). Comparison of 4 525 label-based immunochromatographic assays for the detection of Escherichia coli O157: H7 in milk. 526 Journal of Dairy Science, 100, 5176-5187. https://doi:10.3168/jds.2017-12554 527 McFarland, J. (1907). Nephelometer: An Instrument for Estimating the Number of Bacteria in 528 Suspensions Used for Calculating the Opsonic Index and for Vaccines. Journal of American 529 Medical Association, 14, 1176-1178. https://doi:10.1001/jama.1907.25320140022001f 530 Merkoçi, A., & Morales-Narváez, E. (2015). Sensitive qualitative bioassay using graphene oxide as 531 analyte revealing agent. European Patent No. EP13188693.9. 532 Morales-Narváez, E.& Merkoçi, A. (2012). Graphene Oxide as an Optical Biosensing Platform. 533 Advanced Materials, 24, 3298-3308. https://doi.org/10.1002/adma.201200373 534 Morales‐Narváez, E., Hassan, A. R., & Merkoçi, A. (2013). Graphene Oxide as a Pathogen‐ 535 Revealing Agent: Sensing with a Digital‐Like Response. Angewandte Chemie International 536 Edition, 52 (51), 13779-13783. https://doi.org/10.1002/anie.201307740. 537 Morales-Narváez, E., Naghdi, T., Zor, E., & Merkoçi, A. (2015). Photoluminescent lateral-flow 538 immunoassay revealed by graphene oxide: highly sensitive paper-based pathogen 539 detection. Analytical Chemistry, 87 (16), 8573-8577. https://doi:10.1021/acs.analchem.5b02383 21 540 Ngwa, G. A., Schop, R., Weir, S., León-Velarde, C. G., & Odumeru, J. A. (2013). Detection and 541 enumeration of E. coli O157:H7 in water samples by culture and molecular methods. Journal of 542 Microbiological Methods, 355, 92, 164–172. https://doi.org/10.1016/j.mimet.2012.11.018 543 Olsen, S. J., Miller, G., Kennedy, M., Higgins, C., Walford, J., McKee, G., Fox, K., Bibb, W., & 544 Mead P. (2002). A waterborne outbreak of Escherichia coli O157:H7 infections and hemolytic 545 uremic syndrome: implications for rural water systems. Emerging Infectious Diseases, 8, 370-375. 546 https://doi:10.3201/eid0804.000218 547 Scallan, E., Hoekstra, R. M., Angulo, F. J., Tauxe, R. V. Widdowson, M. A., Roy, S. L., Jones, J. 548 L., & Griffin, P. M. (2011). Foodbome illness acquired in the United States - major pathogens. 549 Emerging Infectious Diseases, 17, 7-15. https://doi:10.3201/eid1701.P11101. 550 Shen, Z., Hou, N., Jin, M., Qiu, Z., Wang, J., Zhang, B., Wang, X., Wang, J., Zhou, D., & Li, J. 551 (2014). A novel enzyme-linked immunosorbent assay for detection of Escherichia coli O157:H7 552 using immunomagnetic and beacon gold nanoparticles. Gut Pathogens, 6, 14. 553 https://doi:10.1186/1757-4749-6-14 554 Shi, X., Meng, X., Sun, L., Liu, J., Zheng, J., Gai, H., Yang, R. & Yeung, E. S. (2010) Observing 555 photophysical properties of quantum dots in air at the single molecule level: advantages in 556 microarray applications. Lab on a Chip, 10, 21, 2844-2847. https://doi:10.1039/c005258b 557 Song, C., Li, J., Liu, J., & Liu, Q. (2016) Simple sensitive rapid detection of Escherichia coli O157: 558 H7 in food samples by label-free immunofluorescence strip sensor. Talanta, 156, 42-47. 559 https://doi.org/10.1016/j.talanta.2016.04.054 560 Tsiraki, M. I., Yehia, H. M., Elobeid, T., Osaili, T., Sakkas, H., & Savvaidis, I. N. (2018). Viability 561 of and Escherichia coli O157: H7 and Listeria monocytogenes in a delicatessen appetizer (yogurt- 562 based) salad as affected by citrus extract (Citrox©) and storage temperature. Food 563 Microbiology, 69, 11-17. https://doi:10.1016/j.fm.2017.07.014 22 564 Zamora-Gálvez, A., Morales-Narváez, E., Romero, J., & Merkoçi, A. (2018). Photoluminescent 565 lateral flow based on non-radiative energy transfer for protein detection in human serum, Biosensors 566 and Bioelectronics, 100, 208-213. https://doi:10.1016/j.bios.2017.09.013. 567 Zhang, L., Huang, Y., Wang, J., Rong, Y., Lai, W., Zhang, J., & Chen, T. (2015). Hierarchical 568 flowerlike gold nanoparticles labeled immunochromatography test strip for highly sensitive 569 detection of Escherichia coli O157: H7. Langmuir, 31, 5537–5544. 570 https://doi:10.1021/acs.langmuir.5b00592 571 Zhaohui, Q., Chunyang, L., Yingchun, Fu., & Yanbin, L. (2017). Rapid and sensitive detection of E. 572 coli O157:H7 based on antimicrobial peptide functionalized magnetic nanoparticles and urease 573 catalyzed signal amplification. Analytical Methods, 9, 5204–5210. https://doi:10.1039/c7ay01643c 574 Zhou, C., Zou, H., Li, M., Sun, C., Ren, D., & Li, Y. (2018). Fiber Optic Surface Plasmon 575 Resonance Sensor for Detection of E. coli O157: H7 based on Antimicrobial Peptides and AgNPs- 576 rGO. Biosensors and Bioelectronics, 117, 347-353. https://doi:10.1016/j.bios.2018.06.005 577 578 579 580 581 582 583 584 585 586 587 588 589 23 590 Figures: 591 592 Figure 1. Fluorescent lateral flow immunoassay (FLFIA) principle and reading. A) FLFIA strip is 593 composed of a sample pad, detection part and an absorbent pad. The detection part of FLFIA strip is 594 composed of a test line TL (Streptavidin-Quantum dots conjugated with biotinylated anti-E. coli 595 O157:H7 antibody “QDs-Ab”) and a control line CL (bare quantum dots “QDs”). When a beef 596 extract or water sample is added to the sample pad of FLFIA strip, it flows by capillary force 597 towards the absorbent pad. If the sample contains E. coli O157:H7, the bacteria will be captured by 598 specific antibody-QDs conjugate on the TL. Afterwards, graphene oxide GO is added onto the 24 599 sample pad. E. coli O157:H7 captured on the TL acts as a spacer between GO and QDs and 600 interrupts the non-radiative energy transfer between them, and this keeps the fluorescence of QDs. 601 On the other hand, the absence of the target bacteria allows the non-radiative energy transfer 602 between GO and QDs on the TL and consequently, quenches the fluorescence of QDs. B) Scanned 603 images with a fluorescence Typhoon reader of two FLFIA strips. Before addition of GO, both TL 604 and CL are fluorescing in both strips. However, after addition of GO, both TL and CL are quenched 605 in negative sample, while, TL of positive sample is still fluorescing. C) Another option of reading 606 the assay is measuring TL and CL intensities by a portable fluorescence lateral flow reader. The 607 measured fluorescence intensities (mV) of both TL and CL, before and after addition of GO to the 608 strip clarify the difference between positive (+) and negative (-) samples. 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 25 625 626 Figure 2. Scanning electron micrographs. A) Heat-killed E. coli O157:H7 (105 CFU mL-1) in 627 standard buffer (10 mM PBS with 0.5 % Tween-20 and 1 % BSA). B) Bare nitrocellulose 628 membrane (Hi-Flow 180, SHF1800425) used for development of detection part of strip. C) GO 629 sheets (150 µg mL-1 with 0.1 % Tween 20) suspended in Milli-Q water, coating E. coli O157:H7 630 cells. C1) A magnified SEM image of E. coli O157:H7 cell is surrounded by GO sheets, C2) Bare 631 GO sheets. D) QDs-anti-E. coli O157:H7 antibody conjugates are capturing to heat-killed E. coli 632 O157:H7 cells in standard buffer. 633 634 635 636 637 638 26 639 27 640 Figure 3. A) Optimization of the quenching process of QDs. Different concentrations of graphene 641 oxide (GO) suspension in Milli-Q water with Tween-20 were investigated to achieve the optimum 642 quenching conditions of both test line (TL) and control line (CL) by measuring the ratio of the final 643 intensity to the initial one (I2/I1) of lines. It shows that GO 150 and 200 μg mL-1 with 0.1 % Tween- 644 20 are the most quenching conditions. B) and C) The significance of using a washing buffer after 645 sample loading to remove nonspecific reaction. B) Without washing step, the initial optimization 646 process exhibited a nonspecific accumulation of the bacterial cells on the CL. The I2/I1 of CL is 647 increasing with bacterial concentration like the TL. C) Conversely, with a washing step, nearly 648 constant I2/I1 of control lines were obtained regardless the bacterial concentrations in the sample. 649 The error bars represent the standard deviation of at least 3 replicates. 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 28 666 29 667 Figure 4. A) Overall response of FLFIA to various concentrations of E. coli O157:H7 in standard 668 buffer. Ratio of test line intensity (RTL)= final intensity/initial intensity of TL. B) Logarithmic 669 response of FLFIA to E. coli O157:H7 (Ec O157:H7) concentrations from 50 to 105 CFU mL-1 in 670 standard buffer. C) Specificity test against higher and lower concentrations of non-specific pathogen 671 (Salmonella Typhimurium, S) either alone or in presence of the target pathogen, E. coli O157:H7 672 (Ec) were investigated. The dashed red lines (A and C) represent the limit of detection of E. coli 673 O157:H7 in standard buffer by FLFIA (≈57 CFU/ mL), which was estimated as the mean value of 674 blank buffer RTL plus three times its SD. The error bars represent the standard deviation of at least 3 675 replicates. 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 30 693 31 694 Figure 5. A) Overall response of FLFIA strips to various concentrations of E. coli O157:H7 (Ec 695 O157:H7) in minced beef and river water samples. The dashed lines represent the limit of detection 696 of E. coli O157:H7 in minced beef (red) and water (blue) (≈178 and 133 CFU g-1 or mL-1 697 respectively), which were calculated as the RTL mean value of blank minced beef or river water plus 698 three times their SD. B) Logarithmic response of FLFIA to E. coli O157:H7 concentrations from 50 699 to 105 CFU g-1 in minced beef. C) Logarithmic response of FLFIA to E. coli O157:H7 700 concentrations from 50 to 105 CFU g-1 in river water. The error bars represent the standard deviation 701 of at least 3 replicates. 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 32 720 Tables 721 Table 1. Spike and recovery experiment in minced beef and river water. Initial level of Spiked value of Real samples RTL in RTL in real Ec O157:H7 Ec O157:H7 Recovery (%) (n ≥ 3) standard buffer samples (CFU mL-1 or g-1) (CFU mL-1 or g-1) River water 0.0 103 0.419 0.411 97.98 0.0 104 0.559 0.532 95.11 Minced beef 0.0 103 0.419 0.398 95.02 0.0 104 0.559 0.501 92.86 722 Where, n, number of replicates. RTL, Ratio of test line intensity= final intensity/initial intensity of 723 test line. Ec O157:H7, E. coli O157:H7. CFU, colony forming units. 724 725 33