Rapid Detection of Foodborne ESBL-Producing Enterobacteriaceae Using MALDI-TOF Mass Spectrometry ^†

Abstract

The proliferation of enterobacteria that produce extended spectrum β-lactamases (ESBLs) has become an important public health concern. The objective of this work was to optimize a phenotypic method for straightforward and rapid ESBL detection in strains obtained from food environments, employing MALDI-TOF mass spectrometry. A collection of 162 enterobacterial isolates, including ESBL producers, AmpC producers, ESBL-AmpC coproducers and sensitive strains was analyzed using a quick method based on cefotaxime hydrolysis. Statistical analysis showed that although the presence of constitutive AmpC hyperproducers could reduce its specificity, this MALDI-TOF assay is an excellent tool for quickly classifying ESBL-producing strains of foodborne Enterobacteriaceae.

1. Introduction

The increasing prevalence of microorganisms producing extended-spectrum (ESBLs) and AmpC β-lactamases poses a critical global challenge in combating antibiotic resistance [1].

Classical approaches for the detection of antimicrobial resistance include disc diffusion and microdilution methodologies, the protocols of which involve isolation and incubation with antibiotic steps, which can delay the acquisition of a result up to three days and can be time-consuming. In clinical settings, obtaining rapid results can be of critical importance in order to make decisions on the treatment of bacterial infections that can be life-threatening. The development of innovative techniques with this objective is therefore essential.

Mass spectrometry for the detection of β-lactamase-producing bacteria has been successfully implemented in clinical laboratories [2], but its application in surveillance and monitoring efforts has not been optimized yet despite its potential to facilitate acquiring data on a bigger scale.

Our objective was to optimize a method for the straightforward and rapid detection of β-lactamase production in strains obtained from environmental and clinical settings employing MALDI-TOF mass spectrometry.

2. Materials and Methods

2.1. Strain Selection

Strains were classified based on the outcome of MAST-D72C (MAST group, UK) and/or microdilution testing in five categories: ESBL positive, constitutive AmpC, inducible AmpC, ESBL + AmpC coproducers and non-beta-lactamase-producing strains.

Table 1. Strain selection of the study. ¹: constitutive AmpC producer; ²: inducible AmpC producer; ³: ESBL + cAmpC coproducer; and ⁴: non-beta-lactamase producing strains.

Species	ESBL	cAmpC 1	iAmpC 2	Coproducer 3	NBLP 4
Escherichia coli	51	0	0	1	35
Escherichia fergusonii	2	0	0	0	0
Enterobacter spp.	0	9	4	1	0
Citrobacter spp.	0	11	10	0	0
Hafnia alveii	0	6	0	0	0
Lelliotia amnigena	0	2	0	0	0
Morganella morganii	0	0	1	0	0
Klebsiella pneumoniae	1	0	0	0	0
Klebsiella oxytoca	1	0	0	1	0
Rahnella aquatilis	10	0	0	0	0
Serratia fonticola	15	0	0	0	0
Scandinavium spp.	0	0	0	1	0
Total	80	28	15	4	35

provides detailed information about the selected strains, which were obtained from various sources, including fresh vegetables, fresh meat, dairy products, wastewater, clinical and collection strains.

2.2. Antibiotic Hydrolysis Assay and Spectrum Acquisition

Test strains were grown on Mueller Hinton Agar (OXOID, Basingstoke, UK) for 18–24 h at 37 °C. A 1 µL loopful of bacterial biomass was resuspended in a 0.5 µg/mL cefotaxime solution in antibiotic reaction buffer (10 mM NH₄CO₃/10 ug/mL ZnCl₂/0.005% SDS) and incubated for 1 h at 37 °C with continuous agitation. After the incubation period, bacterial suspensions were centrifuged at 14,000 rpm for 3 min. Equal parts of the obtained supernatant were mixed with HCCA matrix (Thermofisher, Whaltham, MA, USA) dissolved in 50 HCN/49.9% H₂O/0.1% TFA solution. A total of 1 µL of this mixture was laid on MALDI Target and allowed to dry, avoiding direct light exposure. Spectra were acquired on a MALDI-TOF MS microflex (Bruker, Billerica, MA, USA) with a 0–1000 Da range. MBT-star-ACS solution was employed for calibration using 243.1, 403.3, 573.3 and 757.3 m/z peaks.

2.3. Spectra Data Processing

Acquired spectra were analyzed using FlexAnalysis v.3.4 software (Bruker, Billerica, MA, USA). The 370–414 ± 0.5 and 456–478 ± 0.5 Da peaks were selected as being representative of the hydrolyzed and non-hydrolyzed forms of cefotaxime, respectively, in accordance with SEIMC indications [3]. Spectra were selected for processing if the signal-to-noise ratio of at least one of the pairs of peaks associated with hydrolyzed or non-hydrolyzed forms of the antibiotic was above 5. The intensity of filtered peaks was used for calculating the logRQ parameter [4] using the following equation:

$$\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{R}\mathrm{Q}=\left(\frac{\mathrm{i}370+\mathrm{i}414}{\mathrm{i}456+\mathrm{i}478}\right)$$

The average logRQ from three replicates was normalized between 0 and 1 values using an Excel datasheet.

$$\mathrm{N}\mathrm{o}\mathrm{r}\mathrm{m}\mathrm{a}\mathrm{l}\mathrm{i}\mathrm{z}\mathrm{e}\mathrm{d} \mathrm{l}\mathrm{o}\mathrm{g}\mathrm{R}\mathrm{Q}=\frac{\mathrm{x} \mathrm{l}\mathrm{o}\mathrm{g}\mathrm{R}\mathrm{Q}-\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{R}\mathrm{Q} \mathrm{m}\mathrm{i}\mathrm{n}}{\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{R}\mathrm{Q} \mathrm{m}\mathrm{a}\mathrm{x}-\mathrm{l}\mathrm{o}\mathrm{g}\mathrm{R}\mathrm{Q} \mathrm{m}\mathrm{i}\mathrm{n}}$$

2.4. Statistical Analysis

Hydrolysis peaks were transformed to logRQ and normalized in a datasheet (Microsoft Excel v.2312). All values were statistically analyzed for obtaining the mean and standard deviation. The Levene statistic was performed for testing the homogeneity of group variances. Robust tests of equality of means were performed (Welch and Brown–Forsythe). The effect of types (strains) on the investigated variable (normalized logRQ) was carried out with the Kruskal–Wallis one-way analysis of variance (ANOVA) and post hoc Tamhane’s test (unequal variances). The IBM SPSS Statistics for Windows, Version 26.0 (IBM Corp., Armonk, NY, USA) program was used for data analysis.

3. Results and Discussion

Figure 1 shows a non-ESBL producer exhibiting peaks at 456 and 458 m/z, corresponding to the non-hydrolyzed form of cefotaxime. An ESBL-producing strain that hydrolyzed cefotaxime consequently produced peaks at 370 and 414 m/z, while the peaks at 456 and 458 m/z diminished.

The MALDI-TOF-based ESBL assay for Enterobacteriaceae strains isolated from food and their production environment yielded a global sensitivity of 100% and a specificity of 94.8% (confidence interval 87.2–98.4%), considering all the different types of tested strains in this study (Figure 2 and

Table 2. Threshold values (normalized logRQ) and their respective sensitivity and specificity for the ESBL-producing isolate detection method. In bold: proposed threshold for ESBL detection.

Threshold Value	Sensitivity (%)	Specificity (%)	AUC	Asymptotic Significance
0.000	100	0.0	0.990	0.000
0.100	100	42.7
0.210	100	94.8
0.231	97.5	95.8
0.331	80.0	96.9

). These parameters were obtained from the ROC curve for an optimal cutoff point where sensitivity was maximized. Our results on food strains are comparable to those reported by Oviaño et al. [5] who validated a similar procedure for the classification of ESBL and AmpC producers in positive blood cultures (clinical isolates). They found that the use of cefotaxime resulted in higher sensitivity. Our approach used a more easily understandable normalization procedure (logRQ values from 0 to 1), which would facilitate its application in other laboratories due to its more intuitive interpretation.

All the ESBL producers were accurately identified as positive with a normalized logRQ threshold > 0.210, but there were five Enterobacteriaceae strains that hyperproduced AmpC (false positives). It is important to emphasize that the non-clinical strains co-producing AmpC and ESBL enzymes were detected as negative using the assessed method (Figure 3).

Hydrolysis products of cefotaxime generated by AmpC beta-lactamases are likely different from those generated by ESBL [6]. As a result, they would exhibit different masses. Efforts to identify and distinguish the peaks associated with these metabolites must be conducted in order to accurately classify AmpC-producing strains, particularly hyperproducers, and to increase the specificity of ESBL detection using this methodology.

This study also sheds new light on how constitutive AmpC producers, which we have associated with AmpC-hyperproducing Enterobacteriaceae strains, can be misidentified as ESBL producers. This fact contributes significantly to the lack of total specificity (Figure 3, Table 2). This unspecificity has also been observed in other similar rapid MALDI approaches which do not differentiate the phenotypic expression of AmpC in their validations [5]. Further studies are required to develop similar rapid methods to discriminate strains that exhibit the cAmpC phenotype, especially the hyperproducers that the assessed MALDI-TOF approach categorized as “ESBL”, as well as those strains that co-produced AmpC and ESBL enzymes.

Environmental ESBL strains are generally more susceptible to antibiotics, and food-related strains are no exception to this assumption. This study focused on the detection of food-related ESBL strains, establishing new thresholds for their detection. Our results indicate that even though the presence of constitutive AmpC hyperproducers could potentially reduce its specificity, this MALDI-TOF assay, based on the cefotaxime hydrolysis, demonstrates excellent performance in quickly classifying ESBL-producing strains of Enterobacteriaceae isolated from non-clinical environments, including foods and production farms.