Molecularly imprinted polymer-based photonic crystal sensor array for the discrimination of sulfonamides

Journal Pre-proof Molecularly imprinted polymer-based photonic crystal sensor array for the discrimination of sulfonamides Zheng-zhong Lin, Lei Li, Guiyu Fu, Zhu-zhi Lai, Ai-hong Peng, Zhi-yong Huang PII:

S0003-2670(19)31483-7

DOI:

https://doi.org/10.1016/j.aca.2019.12.032

Reference:

ACA 237315

To appear in:

Analytica Chimica Acta

Received Date: 10 October 2019 Revised Date:

11 December 2019

Accepted Date: 12 December 2019

Please cite this article as: Z.-z. Lin, L. Li, G. Fu, Z.-z. Lai, A.-h. Peng, Z.-y. Huang, Molecularly imprinted polymer-based photonic crystal sensor array for the discrimination of sulfonamides, Analytica Chimica Acta, https://doi.org/10.1016/j.aca.2019.12.032. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. © 2019 Published by Elsevier B.V.

conceptualization: Zheng-zhong Lin methodology: Zheng-zhong Lin software: validation: Guiyu Fu formal analysis: Lei Li; Ai-hong Peng investigation: Lei Li; Guiyu Fu resources: Zhu-zhi Lai data curation: writing (original draft preparation): Lei Li writing (review and editing): Zheng-zhong Lin visualization: supervision: Zhi-yong Huang project administration: funding acquisition:

1

Molecularly imprinted polymer-based photonic crystal sensor array for the

2

discrimination of sulfonamides

3

Zheng-zhong Lin1, Lei Li1, Guiyu Fu1, Zhu-zhi Lai2, Ai-hong Peng1, Zhi-yong

4

Huang1,3,*

5

1. College of Food and Biological Engineering, Jimei University, Xiamen, 361021, China

6

2. College of Chemistry and Chemical Engineering, Xiamen University, Xiamen, 361005, China

7

3. Fujian Provincial Key Laboratory of Food Microbiology and Enzyme Engineering, Xiamen, Fujian Province, 361021, China

8 9

Abstract:

10

In this paper, molecular imprinting and photonic crystal techniques were combined to construct

11

a four-channel sensor array for the simultaneous identification of various sulfonamides. The assay

12

was composed of four units. Three of these units were prepared using sulfaguanidine,

13

sulfamethazine, or sulfathiazole as template molecules. The fourth unit was prepared without a

14

template molecule. The preparation was optimized to obtain maximum identification with a molar

15

ratio of template, monomer, and cross-linker of 1:50:10. The response time was as short as 10 mins.

16

For demonstration, six sulfonamides were selected as analytes. The Bragg diffraction patterns of

17

analytes at different concentrations were measured using the sensor array. Data obtained were

18

analyzed using linear discrimination analysis (LDA) and principal component analysis (PCA). LDA

19

can be applied for SAs discrimination. The message ratios of 87.6%, 94.4%, and 95.8% for six SAs

20

at 10-4 mol L-1, 10-6 mol L-1, and 10-8 mol L-1 were achieved using LDA. The sensor array identified

*

Corresponding author,Zhengzhong Lin, Tel.: [email protected]; Zhiyong Huang, Tel.: [email protected] 1

+86-592-6181912;fax: +86-592-6181912;fax:

+86-592-6180470.E-mail: +86-592-6180470.E-mail:

21

the mixture containing various SAs with an LDA coefficient of 86.1%, thereby indicating that the

22

sensor array had a strong anti-interference ability. The sensor array was used to identify six SAs in

23

fish samples. The measured data in spiked samples were consistent with the fingerprint collected

24

from standard solutions. The accuracy rate reached 90.9%, indicating that the array can be used to

25

identify SAs from food samples.

26

Keywords: Photonic crystal; Molecular imprinting; Sulfonamides; Sensor array

27

1. Introduction

28

Sulfonamides (SAs) are a class of drugs with aminophenylsulfonamide chemical structure.

29

SAs are widely used in animal husbandry and aquaculture, because they can prevent or treat

30

bacterial infections. This kind of drug has many advantages, such as broad antibacterial spectrum,

31

cheap price, and high stability [1]. However, long-term consumption of SAs may accumulate

32

residues in the human and animal body, which may cause diseases, such as liver poisoning, urinary

33

stones, kidney failure, or allergic reactions [2]. Various SAs have different maximum residue limits

34

due to their potential hazards. For most SAs, the maximum residue limit is 0.1 mg/kg. Sulfathiazole

35

and sulfaguanidine are non-detectable. However, SA residues in aquatic products are frequently

36

reported because of the non-standard use of SAs in the breeding process [3]. Currently, the methods

37

used to detect SAs in food are based on high performance liquid chromatography (HPLC), such as

38

HPLC-MS [4], HPLC-UV[5], HPLC-DAD [6], HPLC-FL [7] and others (MS= mass spectrometry,

39

UV= UV visible, DAD= diode array，FL= fluorescence) owing to their advantages, such as high

40

sensitivity and accuracy. These methods have steep requirements, such as high cost, large

41

instruments, and relatively long detection processes. Enzyme-linked immunosorbent assay (ELISA)

42

is a common technique used to detect SAs. ELISA requires antibodies from certain SAs, but

2

43

antibodies are unstable and difficult to prepare. Therefore, it is necessary to develop a rapid method

44

suitable for the screening of SA residues from several sample batches.

45

Molecular imprinting (MI) is a well-established technique through which molecularly

46

imprinted polymers (MIPs) are synthesized and applied with specific shape, size, and binding sites

47

for the certain substance, especially template molecules. The imprinted cavities in MIPs can act as

48

synthetic receptors and exhibit high selectivity toward template molecules [8]. The specific

49

response of MIPs after adsorption of templates can be obtained through optical or electrical signal.

50

MIPs are widely applied to the detection of drug residues in food due to its simple preparation, high

51

efficiency, and reusability [9,10].

52

The present MIP sensors generally adopt visible light, fluorescence, or chemiluminescence as

53

its signal resource. Using this strategy, fluorescence or chemiluminescence groups should be

54

designed and attached to the MIP skeletons, thereby increasing preparation difficulty and lowering

55

universality. In recent years, a kind of sensor called photonic crystal molecularly imprinted polymer

56

(PCMIP), which combines MIP and photonic crystal, was developed to obtain self-reporting

57

(label-free) sensor signal. The adsorption of template molecules by MIP corresponds to change in

58

lattice parameters or refractive index of photonic crystal. The diffraction peak of the photonic

59

crystal, which depends on lattice parameters and refractive index, also shifts. Through structure and

60

size modification, diffraction peaks of PCMIPs vary across the visible light region to attain visual

61

detection. Furthermore, PCMIP sensors exhibit high sensitivity, selectivity, repeatability, stability,

62

fast-response and have been used in the detection of organic molecules, metal ions, and even

63

biomacromolecules. Wang et al. [11] used a PCMIP sensor to detect doxycycline, tetracycline, and

64

aureomycin from milk and honey. Zhao et al. [12] obtained colorimetric measurement of bovine

65

hemoglobin. Hu et al. [13] achieved super-sensitive detection of ephedrine/pseudoephedrine and 3

66

theophylline/caffeine in urine. Wu et al. [14] prepared a PCMIP for visual detection of atrazine with

67

an LOD of 10-16 mol/L. We developed a PCMIP sensor for the analysis of sulfaguanidine in fish

68

samples with a LOD of 2.8×10-10mol L-1 [15]. The linear range was from 10-8 mol L-1 to 10-3 mol

69

L-1. Zhang et al. [16] used sulfamerazine or sulfamethazine as imprinting molecules to prepare a

70

PCMIP sensor with a linear range of 3.8 µM to 22.8 µM. The sensor was applied successfully to

71

detect sulfonamides from egg white samples. Chen et al. [17] developed a PCMIP sensor for the

72

rapid detection of sulfamethazine. The debye-ring diameter of this sensor increased with increasing

73

sulfamethazine concentration. However, the debye-ring diameter was estimated with the naked

74

eyes.

75

MIPs usually show specificity at different extents towards the template and its structural

76

analogs. Sensor arrays can be established based on this MIP characteristic. Cross-reactive sensor

77

arrays have become a powerful tool for the identification and classification of various analytes

78

through distinct patterns [18]. In the construction of the sensor array, a suitable signal source is

79

necessary. The introduction of fluorescence group is a common strategy. Greene [19] integrated

80

seven MIPs and blank polymers into an eight-channel array, which adopted UV-Vis, to distinguish

81

six different aryl amines and their diastereomers. This was the first MIP-based array. They [20] also

82

introduced benzofuran dyes to construct an eight-channel MIP array and used dye substitution

83

method to perform color recognition for seven aromatic amines. Tan et al. [21] used a three-channel

84

fluorescence MIP sensor array to identify five metal ions. Qiu et al. [22] used a three-channel

85

chemiluminescent graphite magnetic MIP sensor array to simultaneously detect diphenol isomers.

86

The adoption of fluorescence group as signal resource usually requires fluorescent labeling,

87

which increases preparation difficulty and cost. Photonic crystal is an ideal signal source, and the

88

sensor array combined photonic crystal and MIP are emerging research subjects. Xu. et al.[23] 4

89

proposed the first PCMIP sensor array, and the team established a four-channel PCMIP sensor array

90

to identify six kinds of contaminants at a wide range of concentrations with 100% accuracy. Also,

91

they [24] established a sensor array, including three PCMIP and one PCNIP elements for the

92

discrimination of trace levels of five poly-brominated diphenyl ethers against a high background of

93

interference with 100% accuracy. However, the research works on PCMIP sensor arrays have

94

seldom been reported in spite of its good performance.

95

Principal component analysis (PCA) is a mathematical process to extract important

96

information from a multivariate data table and to express these information as a set of few new

97

variables called principal components. PCA reduces dimensionality of a multivariate data to two or

98

three principal components that can be visualized graphically, with minimal loss of information.

99

Linear discriminant analysis (LDA) is a type of linear combination, a mathematical process that

100

uses various data items and applies functions to separately analyze multiple classes of objects or

101

items. PCA and LDA are commonly used to analyze a dataset collected from a sensor array to

102

obtain discrimination results for various analytes [23-25].

103

Based on a previous study [15], our group determined a PCMIP prepared from a distinct SA

104

template exhibited diffraction signals varying with different SAs. This finding proved that the

105

adoption of PCMIPs as array elements for SAs discrimination is feasible. In this paper, a

106

four-channel PCMIP sensing array was constructed, and the sensor data for different SAs were

107

collected. The data were analyzed through PCA and LDA analysis to achieve the discrimination of

108

six SAs. This study is the first example of using the PCMIP array for the discrimination of SAs. The

109

constructed array could also be applied for the discrimination of SAs in food samples with high

110

accuracy.

5

111

2. Experimental

112

2.1 Reagents and materials

113

Sulfaguanidine

(SG),

sulfathiazole

(ST),

sulfamethazine

(SM2),

sulfadiazine

(SD),

114

sulfadimethoxine (SDM), and sulfanilamide (SA) were obtained from Shanghai Macklin

115

biochemical Co., Ltd. (Shanghai, China). Azobisisobutyronitrile (AIBN) was purchased from

116

Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China). Tetraethoxysilane (TEOS), ethylene

117

glycol dimethacrylate (EGDMA), and methacrylic acid (MAA) were purchased from Aladdin

118

Reagent Co., Ltd. (Shanghai, China). Ethanol, methanol, and acetic acid were purchased from

119

Xilong Chemical Co., Ltd. (Guangdong, China). All chemicals were analytical grade and were

120

directly used without further purification. The water used in all experiments was purified using

121

TKA Ultra (18.2 MΩ cm, Germany).

122

2.2 Apparatus

123

Diffraction spectra were measured on a fiber optic spectrometer (MAX2000-Pro, Wyoptics Co.

124

Ltd., Shanghai, China). Absorbance spectra were measured using an ultraviolet-visible

125

spectrophotometer (UV-Vis) (Lambda 265, South East Chemicals & Instruments Ltd., Hong Kong,

126

China). The drying of products was carried out using vacuum dryer (DZF-6030, The CIMO

127

Medical Instrument Manufacturing Co., Ltd., Shanghai, China) with a turbine pump (2RK-2B,

128

Boerkang Vacuum Electronics Co., Ltd., Shanghai, China). Microwave heating was conducted

129

using microwave digestion system (MDS-10, Sineo, Microwave Chemistry Technology Co., Ltd.

130

Shanghai, China). Scanning electron microscopy (SEM) images were recorded using JEM-2100

131

microscope (JEOL, Japan) operating at an accelerating voltage of 200 kV. The pH values were 6

132

measured using a pH510 meter (EUTECH, Shanghai, China). Photonic crystal was grown in an

133

incubator with constant temperature and humidity (Yiheng Science Instrument Co. Ltd., Shanghai,

134

China).

135

2.3 Synthesis of disperse SiO2 microsphere

136

SiO2 microspheres with uniform particle size were synthesized using the improved Stöber

137

method [26]. We added 37.5 mL of ethanol and 8.75 mL of ammonia into a 100-mL round-bottom

138

flask. The mixture was sealed with a rubber stopper and stirred magnetically for 5 mins at 30 °C for

139

full mixing. Then, 2.35 mL of TEOS was added into the reaction system at a stirring speed of 1,100

140

rpm within 1 min. Stirring speed was then lowered for enhancing formation of SiO2 microspheres.

141

The product was washed with ethanol and centrifuged trice. Finally, the product was vacuum dried

142

for 6 h at 40 °C. SiO2 microspheres with different particle sizes were obtained by adjusting the ratio

143

of the reactants.

144

2.4 Preparation of photonic crystal film

145

Photonic crystal (PC) film was prepared by vertical deposition method. The dried SiO2

146

microspheres were mixed with ethanol to achieve a mass fraction of 0.5%, which was fully

147

dispersed under ultrasound and then poured into a container. Several glass slides previously cleaned

148

using 7:3 H2SO4:H2O2 (v/v) solution were placed vertically in the container. The container was

149

placed into an incubator at a constant temperature of 40 °C and a constant humidity of 60% for 5

150

days to allow the spontaneous growth of the photonic crystal film. Thus, SiO2 particles

151

systematically self-assembled on the glass slide surface with the natural evaporation of ethanol.

152

2.5 Synthesis of PCMIP sensor array 7

153

Three PCMIPs were prepared using thermal-initiated polymerization method with different

154

templates molecules of SAs [27]. SG was used as an example to illustrate the preparation process.

155

SG at 0.072 mmol was dissolved in 1 mL of methanol, which was then mixed with 346 µL of MAA

156

and 154 of µL EGDMA. The mixture was placed in a refrigerator at 4 °C for 12 h to facilitate the

157

formation of hydrogen bonds between MAA and SG template molecules. AIBN at 0.01 g was then

158

added as a polymerization initiator. The mixture was bubbled with nitrogen for 10 mins for oxygen

159

removal to obtain a pre-polymerization solution. A photonic crystal film was stacked with a

160

polymethyl methacrylate (PMMA) plate to form a sandwich-plate. The sandwich-plate was dipped

161

into the pre-polymerization solution, and the solution was absorbed into the gaps between SiO2

162

microspheres under capillary action. The sandwich-plate was placed in a sealed container for

163

polymerization for 6 h at 60 °C. The imprinted polymer produced was distributed among the gaps

164

between the SiO2 microspheres. Then, the sandwich-plate was treated with HF solution (1%) for 5

165

mins to peel off the glass plate. The SiO2 microspheres and the imprinted polymer were left on the

166

PMMA plate. The PMMA plate was treated with 1% HF solution for another 6 h to remove the SiO2

167

microspheres, thereby creating a polymer film with an inverse opal structure. The SG molecules in

168

the polymer film were eluted with methanol and acetic acid solution (v:v=18:1) until the SG in the

169

elution cannot be detected through UV-Vis. The obtained SG-PCMIP was stored in a buffer solution

170

of NaH2PO4/Na2HPO4 (0.01 mol L-1, pH 7.6). The SM2-PCMIP and ST-PCMIP were prepared

171

following the same method as SG-PCMIP but using SM2 and ST as template molecules. In addition,

172

a PCNIP film was prepared using the same method but without any template molecules. The

173

detailed experimental scheme was illustrated in Figure 1. Experimental conditions, such as solvent,

174

ratio of template to monomer, and ratio of template to cross-linker, were optimized because they

175

affected the sensitivity of the sensor and the discrimination ability of the array [14-16]. These 8

176

conditions were optimized based on the signal response value and by changing the various solvents

177

and ratios with single factor analysis method.

178

2.6 Diffraction spectra measurement

179

In a typical assay, PCMIP and PCNIP films were respectively immersed in 20 mL of

180

NaH2PO4/Na2HPO4 buffer solutions (0.01 mol L-1, pH 7.6) containing SG, SM2, or ST at different

181

concentrations with an incident angle of 0°. After incubation for 10 mins at room temperature, the

182

intensity and wavelength of diffraction peaks of PCMIP and PCNIP films were measured using a

183

fiber optic spectrometer.

184

2.7 Principal component analysis and linear discriminant analysis

185

PCA is a multivariate statistical analysis method that selects fewer important variables through

186

linear transformation of multiple variables to reduce the dimension of data [23,24]. LDA is a

187

multivariate statistical analysis method for identification [25]. Generally, LDA is expected to have

188

an identification or group function superior to that of PCA.

189

In this experiment, the data in five parallel tests about diffraction peak shift values (∆λ)

190

against six sulfonamide substances at three concentrations (10-4 mol L-1, 10-6 mol L-1, and 10-8 mol

191

L-1) were collected from four sensor units (3 PCMIP and 1 PCNIP), which were analyzed through

192

PCA and LDA using SPSS Statistics 19 software.

193

3. Results and Discussion

194

3.1 Particle size selection

9

195

According to the Bragg equation (λmax=1.633D(n2avg - sin2θ)1/2), the wavelength of diffraction

196

peak of photonic crystal has a linear relationship with lattice constant. In this experiment, SiO2

197

microspheres with different particle sizes were prepared using the Stöber method. These

198

microspheres were characterized using SEM and particle size statistical analysis. The results are

199

shown in Figure 2. The particle sizes of SiO2 prepared were centered at 246, 330, 375, 450, and 625

200

nm, respectively (Figure 2A-2E). The SEM diagram of PC membrane prepared with SiO2

201

microspheres (375 nm) is shown in Figure 2F, in which the periodical structure was clearly

202

observed. The corresponding reflection peaks of photonic crystal film were measured at 510, 600,

203

660, 775, and 890 nm, as shown in Figure 3A. The wavelengths of diffraction peaks were basically

204

proportional to the particle sizes (Figure 3B), thereby implying that the SiO2 microspheres in

205

photonic films were arranged in a periodical array, as shown in Figure 2. The wavelengths

206

diffraction peaks of photonic crystals are affected by the SiO2 microsphere arrangement, which is

207

dependent on the growth environment. Therefore, the key to this experiment is to control the growth

208

environment and the size of SiO2 microspheres. By comparing the diffraction intensity and peak

209

shape of five photonic crystal films (A-E in Figure 3A) under different growth environments, C and

210

D resulted in in high intensity and narrow half-peak width, and the PCMIP prepared with C was

211

more stable than D. Therefore, the particle size of SiO2 microspheres was controlled at 375 nm.

212

3.2 Optimization of preparation conditions of PCMIPs

213

Generally, the optimal index for a single sensor is its response value for high sensitivity. In this

214

case, high wavelength shift value (△λ) indicated that the sensor unit had high responsibility for the

215

analyte. However, the sensor array was used to perform a thorough discrimination of analytes.

216

Therefore, the optimal index was chosen to obtain the maximum difference between wavelength 10

217

shifts from three sensor units for their corresponding sulfonamide template molecules to ensure

218

highest identification of different SAs [22].

219

Reaction solvent plays an important role in MI [28]. An ideal solvent has sufficiently high

220

solubility for template molecules and does not or slightly interfere with the interaction between

221

functional monomers and template molecules. The optimization result of solvents showed that

222

acetonitrile has the highest shift value (Figure S1a) and difference (Figure S1b), whereas methanol’s

223

performance was a bit lower than acetonitrile. However, acetonitrile has corrosive effect on PMMA

224

plate to an extent that would damage the integrity of PCMIP film. Comprehensive consideration

225

suggested that methanol was the best solvent.

226

As shown in Figure S2, the molar ratio of MAA and EGDMA was optimized in this experiment.

227

Excessive EGDMA enabled the formation of a rigid framework structure of PCMIP film to ensure

228

sharp diffraction peak with desensitization. Alternatively, excessive MAA was beneficial for high

229

flexibility and sensitivity of PCMIP film, but it caused a vulnerable framework with poor diffraction

230

peak [29]. The molar ratio of MAA to EGDMA of 5:1 had the highest shift value (Figure S2a) and

231

difference (Figure S2b). The PCMIP film prepared at the ratio showed moderate rigidity and

232

flexibility.

233

The ratio of template molecule to functional monomer produced a distinct effect on the

234

adsorption amount of PCMIP films [30]. As shown in Figure S3 at a molar ratio of 1:50, the

235

differences for the three PCMIPs that reached the maximum and the shift values were the highest

236

(SG-PCMIP) or the second higher (SM2 and ST-PCMIP). In addition, SM2 and ST cannot be fully

237

dissolved at ratios exceeding 1:50. Thus, the optimal ratio of template molecule to functional

238

monomer was 1:50.

11

239

In conclusion, the conditions were optimized as the molar ratio of SAs, MAA, and EGDMA

240

(1:50:10) with methanol as solvent.

241

3.3 Characterization of PCMIPs

242

According to the SEM images in Figure 4A, SiO2 microspheres were evenly and orderly

243

arranged on the surface of the glass slide, forming an opal structure with symmetry of face-centered

244

cube. According to particle size statistics, the average particle size of SiO2 microspheres was

245

approximately 375 nm with a diffraction peak at 660 nm. In Figure 4B, the PCMIPs film presented

246

an inverse opal structure packing in a manner of ABAB…. Intersecting cavities caused by the

247

elution of SiO2 microspheres were observed. This multi-layer porous structure had high surface area,

248

which facilitated the transfer of target molecules among the PCMIP film. The average period of the

249

PCMIPs film was measured at approximately 270 nm in length from SEM images. The

250

corresponding diffraction peak located at 550 nm from diffraction spectrum is shown in Figure 4C.

251

The blue shift of diffraction peak indicated that the size of cavities in PCMIP was smaller than SiO2

252

microspheres in PC, and such difference was caused by framework contraction after elution of SiO2

253

microspheres.

254

3.4 Response time

255

The kinetic experiments of sensor units for their template SAs were performed to determine

256

their equilibrium response time. The concentrations of all the SAs were 10-4 mol L-1. The kinetic

257

curves were shown in Figure 5. The three PCMIPs had the similar response curves. The peak shift

258

values noticeably increased and became stable. The equilibrium response time was hypothesized to

259

be 10 min. The peak shift value of SG PC-MIP was the highest and SM2 PC-MIP was the lowest at 12

260

similar response time, which may have arisen from different molecular structure and size of SAs.

261

SM2 was the SA with the most complex molecular structure, whereas SG was the simplest among

262

the three SAs, according to the response sequence in the kinetics experiments.

263

3.5 Responses of sensor units

264

The sensor array, constructed from four sensor units, SG-PCMIP, SM2-PCMIP, ST-PCMIP, and

265

PCNIP, was used to collect signals from six SAs (SG, SM2, ST, SA, SDM, and SD) at

266

concentrations of 10-4, 10-6, and 10-8 mol L-1 with five parallel tests to obtain their corresponding

267

fingerprints. The diffraction patterns of SG-PCMIP, SM2-PCMIP, and ST-PCMIP of the six SAs at

268

the concentrations of 10-4 mol L-1 were shown in Figure 6A. The SG-PCMIP, SM2-PCMIP, and

269

ST-PCMIP exhibited the highest signal response toward their template molecules, such as SG, SM2,

270

and ST, respectively. The three PCMIPs had abundant imprinted sites for their corresponding

271

template molecules. Thus, the PCNIP had the lowest shift values and the slightest signal difference

272

for six SAs than any one of the PCMIPs, probably because of the absence of imprinted sites for SAs

273

in PCNIP. Each sensor unit had its individual signal set for the six SAs to construct a set of

274

fingerprints that can distinguish the six SAs. The sensor array exhibited the similar fingerprint set at

275

the concentrations of 10-6 mol L-1 and 10-8 mol L-1, as shown in Figure S4A and S5A, which

276

indicated that the sensor array was capable to distinguish six SAs in a wide concentration range.

277

3.5 Principal component analysis

278

PCA was used to reduce the above multi-dimensional data to 2D or 3D dataset [23, 24]. PCA

279

reduced the complexity and size of the training data (4 polymers–6 analytes–5 replicates) and

280

transformed them into roots that were linear combinations of the response patterns. Based on the

281

above signals collected from four sensor units at a concentration of 10-4 mol L-1, two principal 13

282

components (PC1 and PC2) were extracted after PCA analysis. The coefficients from PC1 and PC2

283

were greater than one, and the cumulative percentage reached 88.0%, which indicated that PC1 and

284

PC2 almost contained the total information of the four variables. According to the statistical law, the

285

sum of information amount for principal components greater than 80% indicated that these principal

286

components can represent all variables [31]. A PCA scatter plot was shown in Figure 6B, in which

287

the signals from different SAs were clustered in different areas. This fact implied that the PCA

288

could discriminate various SAs. The signals from another two concentrations at 10-6 and 10-8 mol

289

L-1 were also treated by PCA and the PCA scatter plots, as shown in Figures S4B and S5B. The

290

signals from different SAs were also distinctly distributed. The percentages of PC1 and PC2

291

decreased and increased, respectively, with the concentration of SAs, thereby indicating that the

292

characteristic values from original information disappeared [32].

293

3.6 Linear discrimination analysis

294

The data matrix obtained from sensor units can likely be intuitively represented through the

295

two-dimensional LDA plot, in which each point represented a measurement result, and each ellipse

296

represented the confidence interval of a certain analyte [33,34]. The responses of the assay for six

297

SAs at the concentration of 10-4 mol L-1 were clustered into six tight groups, which were

298

individually separated in the LDA plot (Figure 6C), indicating that the sensor assay exhibited

299

excellent discrimination for the six SAs. The extracted two functions, LD1 and LD2, occupied 87.6%

300

of the total information. This value was a bit lower than PCA (88.0%) at a similar concentration.

301

However, the data in a certain ellipses of LDA were distributed more closely than PCA, thereby

302

indicating that LDA provided discrimination results better than PCA. Similarly, Figures S4C and

303

S5C showed the LDA plots at lower SA concentrations (10-6 and 10-8 mol L-1). The responses of the 14

304

assay for six analytes were distinctly classified with 94.4% and 95.8% of the total information. The

305

extracted information percentages were over 85% at different concentrations, which implied that the

306

array could discriminate various SAs even at low concentrations.

307

LDA is a supervised method for separating classes of objects and for assigning new

308

components to appropriate classes [25]. The LDA results exhibited a satisfying classification

309

performance, which achieved 100% accuracy for both original groups and cross-validation

310

procedure for all three concentrations.

311

The application of the array for the sensor of complex solutions, such as mixtures of analogues

312

is crucial [23-25]. To demonstrate, six SAs were grouped according to their positions in the LDA

313

plots, two neighboring SAs in the LDA plots were assigned in a group, as follows: SG and SA were

314

in group 1; SM2 and SDM were in group 2; and ST and SD were in group 3. They were

315

discriminated by using the LDA method. The result is shown in Figure 7A. Each group was

316

represented by an individual ellipse in the plot. The extracted two functions, namely, LD1 and LD2,

317

occupied as high as 100% of the total information. In group 1, SG and SA had close molecular

318

structure with small size and steric hindrance. In group 2, SM2 and SDM only had a slight

319

difference in molecular structure (–CH3 and –OCH3). In group 3, ST and SD had similar structures.

320

Thus, the array exhibited similar signal responses for SAs in the same group. However, even for the

321

two SAs in one group (ellipse), distinct distribution difference was also observed, indicating that the

322

array was able to discriminate SAs with tight structure differences.

323

The discrimination experiments of the array for SA mixtures were also performed to

324

investigate the anti-interference ability of the approach in the presence of other SAs. Four solutions

325

were chosen, as follows: single analyte of SG; two-analyte mixture of SG + SM2; three-analyte

326

mixture of SG+SM2+ST; and four-analyte mixture of SG+SM2+ST+SA. The discrimination result 15

327

is shown in Figure 7B. The extracted functions, LD1 and LD2, occupied 86.1% of the total

328

information, thereby indicating that the array achieved the discrimination for 4 mixtures with

329

growing interfering substances.

330

3.7 Sample analysis

331

The application of the array for sensing complex solutions, such as food sample is a huge

332

challenge. The practical application of the array was investigated through SA residues’

333

discrimination in fish. The results were shown in Figure 8. The solid ellipses of 1-6 represented the

334

discrimination of the spiked SAs, and the dotted ellipses represented the predicted areas at which

335

the spiked SAs should be. Clearly, the expected areas and the measured areas matched well with a

336

discrimination accuracy of 90.9% in spite of the interference from complex components in the fish

337

matrix. Food analysis using MIP arrays is seldom reported. Thus, the as-prepared array proposed a

338

strategy applicable in food analysis.

339

4. Conclusions

340

A sensor array combining MI technology and optical self-reporting was prepared. The array

341

was composed of four kinds of sensor units that had various diffraction signals against six SAs. The

342

response time was only 10 mins. A molar ratio of 1:50:10 of template, monomer, and cross-linker

343

was proposed based on the optimum experiments. The signals of the array for six SAs were

344

collected and processed using PCA and LDA methods. The array was confirmed to discriminate six

345

individual SAs or their mixtures at three concentrations. Moreover, the array can be applied to the

346

SA discrimination of fish samples with an accuracy of 90.9%.

347

Acknowledgements

16

348

This research was supported by the Natural Science Foundation of Fujian Province of China

349

(2018J01432, 2017J01633), the Open Foundation from Key Laboratory of Food Microbiology and

350

Enzyme Engineering of Fujian Province (B18097-5), cultivation project in Jimei University for

351

national foundation (ZP2020029), and the National Undergraduate Training Programs for

352

Innovation and Entrepreneurship (201910390017, 201810390024, 201710390300).

353

Interest statement

354

Declaration of interest: none

355

References

356

[1] Z.G. Chen, C.R. Zhan, P. Guo, Y.X. Wang, H.U. Zhi-Guo, Study on Simultaneous Determination of 13

357

Sulfonamides Residues in Aquatic Products by HPLC, Food Sci. 28 (2007) 448-451.

358

[2] H. Khalili, A. Soudbakhsh, A.H. Talasaz, Severe hepatotoxicity and probable hepatorenal syndrome associated

359

with sulfadiazine, Am. J. Health-Syst. Ph. 68 (2011) 888-892.

360

[3] C. Blasco, Y. Picó, C.M. Torres, Progress in analysis of residual antibacterials in food, Trend Anal. Chem. 26

361

(2007) 895-913.

362

[4] C. Nebot, P. Regal, J.M. Miranda, C. Fente, A. Cepeda, Rapid method for quantification of nine sulfonamides

363

in bovine milk using HPLC/MS/MS and without using SPE, Food Chem. 141 (2013) 2294-2299.

364

[5] M. Karimi, F. Aboufazeli, H.R.L.Z. Zhad, O. Sadeghi, E. Najafi, Determination of sulfonamides in chicken

365

meat by magnetic molecularly imprinted polymer coupled to HPLC-UV, Food Anal. Method. 7 (2014) 73-80.

366

[6] R. Qi, X. Lv, N. Qian, B. Hu, Q. Jia, Rapid HPLC-DAD quantitation of sulfonamides in honey using

367

poly(methacrylic

368

methacrylate-grafted sodium titanate nanotubes, New J. Chem. 39 (2015) 6323-6331.

369

[7] A. Zotou, C. Vasiliadou, LC of sulfonamide residues in poultry muscle and eggs extracts using fluorescence

370

pre‐column derivatization and monolithic silica column, J. Sep. Sci. 33 (2015) 11-22.

acid-ethylene

dimethacrylate)

monolith

17

modified

with

3-(trimethoxysilyl)propyl

371

[8] L. Chen, S. Xu, J. Li, Recent advances in molecular imprinting technology: current status, challenges and

372

highlighted applications., Chem. Soc. Rev. 40 (2011) 2922-2942.

373

[9] Y. Zhao, L. Zhou, S. Pan, P. Zhan, X. Chen, M. Jin, Fast determination of 22 sulfonamides from chicken

374

breast muscle using core–shell nanoring amino-functionalized superparamagnetic molecularly imprinted polymer

375

followed by liquid chromatography-tandem mass spectrometry, Journal of Chromatography A 1345 (2014) 17-28.

376

[10] Z. Lin, H. Zhang, L. Li, Z. Huang, Application of magnetic molecularly imprinted polymers in the detection

377

of malachite green in fish samples, React. Funct. Polym. 98 (2016) 24-30.

378

[11] L. Wang, F. Lin, L. Yu, A molecularly imprinted photonic polymer sensor with high selectivity for

379

tetracyclines analysis in food, Analyst 137 (2012) 3502-3509.

380

[12] Y.J. Zhao, X.W. Zhao, J. Hu, J. Li, W.Y. Xu, Z.Z. Gu, Multiplex Label-Free Detection of Biomolecules with

381

an Imprinted Suspension Array, Angew. Chem. 48 (2010) 7350-7352.

382

[13] X. Hu, G. Li, M. Li, J. Huang, Y. Li, Y. Gao, Y. Zhang, Ultrasensitive specific stimulant assay based on

383

molecularly imprinted photonic hydrogels, Adv. Funct. Mater. 18 (2008) 575-583.

384

[14] Z. Wu, C.A. Tao, C. Lin, D. Shen, G. Li, Label-free colorimetric detection of trace atrazine in aqueous

385

solution by using molecularly imprinted photonic polymers, Chemistry 14 (2010) 11358-11368.

386

[15] L. Li, Z. Lin, Z.Y. Huang, A.H. Peng, Rapid detection of sulfaguanidine in fish by using a photonic crystal

387

molecularly imprinted polymer, Food Chem. 281 (2019) 57-62.

388

[16] Y. Zhang, H. Ren, L. Yu, Development of molecularly imprinted photonic polymer for sensing of

389

sulfonamides in egg white, Anal. Methods-UK 10 (2018) 101-108.

390

[17] X. Chen, G. Liu, C. Ren, M. Gao, X. Fan, Investigation on Sulfamethazine Molecularly Imprinted

391

Two-dimensional Photonic Crystal Hydrogel Sensor, Chemical Journal of Chinese Universities 39 (2018)

392

212-218.

18

393

[18] A. Bajaj, O.R. Miranda, I.B. Kim, R.L. Phillips, V.M. Rotello, Detection and differentiation of normal,

394

cancerous, and metastatic cells using nanoparticle-polymer sensor arrays, P. Natl. Acad. Sci. USA. 106 (2009)

395

10912-10916.

396

[19] N.T. Greene, S.L. Morgan, K.D. Shimizu, Molecularly imprinted polymer sensor arrays, Chem. Commun.

397

(2004) 1172-1173.

398

[20] N.T. Greene, K.D. Shimizu, Colorimetric molecularly imprinted polymer sensor array using dye

399

displacement, J. Am. Chem. Soc. 127 (2005) 5695-5700.

400

[21] J. Tan, H. Wang, X. Yan, A fluorescent sensor array based on ion imprinted mesoporous silica, Biosens.

401

Bioelectron. 24 (2009) 3316-3321.

402

[22] H. Qiu, C. Luo, M. Sun, F. Lu, L. Fan, X. Li, A chemiluminescence array sensor based on

403

graphene-magnetite-molecularly imprinted polymers for determination of benzenediol isomers, Anal. Chim. Acta

404

744 (2012) 75-81.

405

[23] D. Xu, W. Zhu, C. Wang, T. Tian, J. Cui, J. Li, H. Wang, G. Li, Molecularly imprinted photonic polymers as

406

sensing elements for the creation of cross-reactive sensor arrays, Chem.-Eur. J. 20 (2015) 16620-16625.

407

[24] D. Xu, W. Zhu, C. Wang, T. Tian, J. Li, Y. Lan, G. Zhang, D. Zhang, G. Li, Label-free detection and

408

discrimination of poly-brominated diphenyl ethers using molecularly imprinted photonic cross-reactive sensor

409

arrays, Chem. Commun. 50 (2014) 14133-14136.

410

[25] F. Deng, W. Chen, J. Wang, Z. Wei, Fabrication of a Sensor Array Based on Quartz Crystal Microbalance

411

and the Application in Egg Shelf Life Evaluation, Sensor. Actuat. B Chem. 265 (2018) 394-402.

412

[26] W. Stöber, A. Fink, E. Bohn, Controlled growth of monodisperse silica spheres in the micron size range, J.

413

Colloid Interf. Sci. 26 (1968) 62-69.

414

[27] Z. Lin, D. Wang, A. Peng, Z. Huang, Y. Lin, Determination of domoic acid in shellfish extracted by

415

molecularly imprinted polymers, J. Sep. Sci. 39 (2016) 3254–3259. 19

416

[28] M.C. Fonseca, C.S.N. Jr, K.B. Borges, Theoretical investigation on functional monomer and solvent selection

417

for molecular imprinting of tramadol, Chem. Phys. Lett. 645 (2016) 174-179.

418

[29] Z. Ateş, O.G. Uuml, Ven, Radiation-induced molecular imprinting of -glucose onto poly(2-hydroxyethyl

419

methacrylate) matrices using various cross-linkering agents, Radiat. Phys. Chem. 79 (2010) 219-222.

420

[30] B. Yu, X. Zhang, H. Jie, K. Yang, C. Zhao, Effect of the template molecules and nonsolvent additives on the

421

recognition property of molecular imprinted polyethersulfone particles, J. Appl. Polym. Sci. 108 (2010)

422

3859-3866.

423

[31] S. Roweis, Em algorithms for pca and spca, Adv. Neural Inf. Process. Syst. 10 (1997) 626-632.

424

[32] P. Chang, Z. Zhang, C. Yang, Molecularly imprinted polymer-based chemiluminescence array sensor for the

425

detection of proline., Anal. Chim. Acta 666 (2010) 70-75.

426

[33] S. Abbasimoayed, H. Golmohammadi, A. Bigdeli, M.R. Hormozinezhad, A rainbow ratiometric fluorescent

427

sensor array on bacterial nanocellulose for visual discrimination of biothiols., Analyst 143 (2018) 3415-3424.

428

[34] S. Liang, C. Liu, M. Watanabe, B. Chen, K. Hayashi, LSPR sensor array based on molecularly imprinted

429

sol-gels for pattern recognition of volatile organic acids, Sensor. Actuat. B Chem. 249 (2017) 14-21.

430 431

Figure Captions

432

Figure 1. Schematic illustration of a four-channel array fabrication based on colloidal-crystal molecular

433

imprinting.

434

Figure 2. A–E: SEM and particle size distribution of SiO2 with different particle sizes (A: 264.72 nm, B: 329.56

435

nm, C: 375.23 nm, D: 449.76 nm, and E: 623.65 nm). F: SEM and particle size distribution of a PC membrane

436

prepared with SiO2 microspheres (375 nm).

437

Figure 3. (A) Reflection peak of PC membrane obtained from SiO2 microspheres with different particle sizes

438

(where S is the SiO2 microspheres at 200 nm purchased for comparison, and A–E are the SiO2 microspheres with 20

439

different particle sizes obtained by adjusting reaction conditions). (B) Linear relationship between the particle size

440

of SiO2 microsphere and the reflection peak position of PC membrane.

441

Figure 4. SEM and particle size distribution of the PC film (A), PCMIP film (B), and their reflection spectra (C).

442

Figure 5. Dynamic response curve of sensor unit (The concentration of SG/SM2/ST is 10-4 mol L-1).

443

Figure 6. Discrimination ability of the sensor array for 6 SAs at 10-4 mol L-1 (n=5). (A) Characteristic shift values

444

of each sensor unit for the six SAs used to perform PCA and LDA. (B) PCA plot at two principal components with

445

a cumulative eigenvalue of 88.0%. (C) LDA plot at two discrimination functions with a cumulative eigenvalue of

446

87.6%.

447

Figure 7. Linear discriminant diagrams (A) for SAs in the three groups (1:SG+SA; 2: SM2 + SDM; 3: ST + SD;

448

C=10-4 mol L-1, n=5) and (B) for SA mixtures (1:SG, 2:SG+SM2, 3:SG+SM2+ST, 4:SG+SM2+ST+SA; C=10-4

449

mol L-1, n=5).

450

Figure 8. Linear discriminant diagrams for SAs in fish samples (n=5, C= 10-4 mol L-1).

451 452

21

Molecularly imprinted polymer-based photonic crystal sensor array for the discrimination of sulfonamides Zheng-zhong Lin1, Lei Li1, Guiyu Fu1, Zhu-zhi Lai2, Ai-hong Peng1, Zhi-yong Huang1,3,*

Figure S1. (a) Wavelength shifts from three sensor units for their corresponding sulfonamide template molecules in various solvents, for example SG-PCMIP vs SG. High wavelength shift value indicates the sensor unit behaves high responsibility for its template.(b) Differences of wavelength shifts from three sensor units in various solvents. High difference indicates that the sensor array (three sensor units) behaves high discrimination ability for different sulfonamides.(DMSO represented dimethyl sulfoxide, MeCN represented acetonitrile, MeOH represented methanol, TM represented trichloromethane). The concentrations of sulfonamides (SG/SM2/ST) were 10-4 mol L-1. The molar ratio of MAA to EGDMA was fixed at 7.5:1 and sulfonamides to MAA was at 1:70.

Figure S2. (a) Wavelength shifts from three sensor units for their corresponding sulfonamide template molecules under different molar ratios of MAA to EGDMA.(b) Differences of

wavelength shifts from three sensor units under different molar ratios of MAA to EGDMA. The concentrations of SG/SM2/ST were 10-4 mol L-1. The solvents were MeOH, The molar ratio of sulfonamides to MAA was at 1:70.

Figure S3. (a) Wavelength shifts from three sensor units for their corresponding sulfonamide template molecules under different molar ratios of SAs to MAA.(b) Differences of wavelength shifts from three sensor units under different molar ratios of SAs to MAA. The concentration of SG/SM2/ST is 10-4 mol L-1. The solvents were MeOH, The molar ratio of MAA to EGDMA was fixed at 7.5:1.

Figure S4. Discrimination ability of the sensor array for 6 SAs at a concentration of 10-6mol L-1. (n=5). (a) Characteristic shift values of each sensor unit for 6 SAs that were used to perform PCA and LDA. (b) PCA plot at two principal components with a cumulative eigenvalue of 88.7%. (c)

LDA plot at two discrimination functions with a cumulative eigenvalue of 94.4%.

Figure S5. Discrimination ability of the sensor array for 6 SAs at a concentration of 10-8mol L-1. (n=5). (a) Characteristic shift values of each sensor unit for 6 SAs that were used to perform PCA and LDA. (b) PCA plot at two principal components with a cumulative eigenvalue of 86.2%. (c) LDA plot at two discrimination functions with a cumulative eigenvalue of 95.8%.

Photonic crystal imprinting array was firstly reported for sulfonamide discrimination The applicable concentration range of sulfonamide is wide (10-4~10-8 mol/L) The response is rapid and label-free The array was applicable in food analysis

Declaration of interests ☐The authors declare that they have no known competing financialinterestsor personal relationships that could have appeared to influence the work reported in this paper. ☐The authors declare the following financial interests/personal relationships which may be considered as potential competing interests: