The following reagents were employed to optimise the LAMP assays for the detection and quantification of lactic acid bacteria: Bst DNA polymerase 2.0 (New England, BioLabs), magnesium sulphate (MgSO₄), 10× isothermal amplification buffer, deoxynucleotide triphosphates (dNTPs) (NZYtech, Lisbon, Portugal), custom-designed primers (Metabion, Munich Germany), hydroxy naphthol blue (HNB) (Metabion, Munich Germany), neutral red (ZellBio GmbH, Blaubeuren Germany), gold nanoparticles (AuNPs) 5 nm citrate-stabilised (Sigma-Aldrich, Tres Cantos, Madrid), EvaGreen dye (20×, BIOTIUM), TEN buffer, TE buffer, and propidium monoazide (PMA; Sigma-Aldrich).

Ten bacterial strains were employed, comprising nine LAB species and one Acetobacter species (Table 1). Additionally, some yeast and other bacteria were used; however, the data are not presented. LAB were cultivated in MRS medium, except Oenococcus oeni L33 and Acetobacter aceti CECT 298^T, which were cultured in MLO medium (Zúñiga et al., 1993). The strains were inoculated into commercial red (Bobal) and white (Chardonnay) wines and corresponding grape musts. Cell concentrations were verified using a haemocytometer, and all experiments were conducted in triplicate.

Cell suspensions from lab media were only washed twice with Milli-Q water; cells from white wine and musts were additionally washed with 10 % TEN buffer (0.1 M Tris-HCl, pH 7.5, 0.05 M EDTA, and 0.8 M NaCl), followed by twice in Milli-Q water and finally suspended in TE buffer. Cell suspensions from red wine and red grape must were washed with 10 % TEN buffer supplemented with 2 % polyvidone (PVP) and twice with Milli-Q water to eliminate phenolic interference. The genomic DNA extraction was achieved by vortexing with 0.5 mm glass beads for 10 minutes (Soares-Santos et al., 2018).

Primers specific to LAB were designed using the Genome-based LAMP Primer Design tool (GLAPD): (http://cgm.sjtu.edu.cn/GLAPD/online/) (Jia et al., 2019), targeting conserved regions of the 16S rRNA gene of lactic acid bacteria. For this purpose, the Lactiplantibacillus plantarum genome sequence was used as a reference genome, and Oenococcus, Lactobacillus, and Pediococcus species were used as the target group in GLAPD. Yeast and some acetic acid bacteria (A. aceti and Gluconobacter oxydans) were used in the background group for LAB primer generation. Each set included two outer primers (F3 and B3) and two inner primers (FIP and BIP). Primer specificity was confirmed via sequence alignment using MEGA X software against reference genomes of LAB and non-LAB strains.

To minimise nonspecific amplification, various concentrations of gold nanoparticles (0, 1, 3, and 5 µL) were evaluated. The optimal condition, yielding the clearest distinction between positive and negative reactions, was determined to be 5 µL per 25 µL LAMP reaction. Reactions were conducted at 62 °C for 30 minutes in a heat block.

Each LAMP reaction mixture (25 µL) comprised: 1.4 mM dNTPs, 0.2 µM each outer primer, 1.6 µM each inner primer, 8 mM MgSO₄, 1× buffer, 5 µL gold nanoparticles, 0.32 U/µL Bst polymerase, 7.5 µL extracted DNA, and sterile Milli-Q water (negative control).

Colorimetric detection was performed using HNB (120 µM) and neutral red (100 µM) dyes, and also by turbidity in the absence of any dye. Colour changes were assessed by visual inspection and with the HNB enhance app (https://colorimetry.net/hnb-app/) to assist interpretation. Additionally, gel electrophoresis (2 % agarose) was used to confirm amplicon formation.

Semi-quantification was achieved by tracking the change in red colour intensity (RGB values) every five minutes during amplification using the “What a color?” app (https://apps.apple.com/us/app/what-a-color/id1229503218). A reduction in red intensity correlated with increasing DNA amplification (Figure 1).

Endpoint measurements were obtained by transferring 20 µL of LAMP product into 364-well microtiter plates. Optical density was recorded using a FLUOstar OPTIMA plate reader (BMG LABTECH) with an excitation filter at 340–10 nm. A standard curve was constructed using LAB suspensions ranging from 10² to 10⁸ CFU/mL in the same matrix as the samples (red or white grape musts and wines).

Quantitative LAMP (qLAMP) was conducted on a StepOne™ real-time PCR system. The reaction mix was identical to the standard LAMP, with the substitution of EvaGreen (1.5 µL) as the fluorescence dye. Amplification curves and melt profiles were generated to assess specificity and quantify LAB concentrations.

To distinguish viable from dead cells, LAB cultures were heat-killed (100 °C for 30 minutes). PMA (50 µM) was added to cell suspensions (6.3 × 10⁶ CFU/mL), incubated in the dark for 15 minutes, then photoactivated using a PhAST Blue LED system (465–475 nm, 30 minutes). PMA-treated and untreated samples were then subjected to qLAMP under the same conditions as above.

Two primer sets targeting conserved regions of the 16S rRNA gene were designed using the GLAPD software. These comprised outer primers LAB-F3 and LAB-B3 and inner primers LAB-FIP and LAB-BIP, optimised for high specificity to LAB. Sequences of the four LAB general primers are: LAB-F3 (AGTGAGTGGCGAACTGGT), LAB-B3 (GCCGATTACCCTCTCAGGT), LAB-FIP (GCGGTCCAAGTTGTTATGCGGTGAGTAACACGTGGGAAACCT), and LAB-BIP (ATCATTTTGGATGGTCCCGCGTGCCATGGTGAGCCGTTA). Sequence alignment using MEGA X confirmed a high identity between the designed primers and the LAB strains employed in this study, while non-target organisms, including A. aceti, showed negligible alignment, as shown in Figure S1. This in silico validation was further corroborated by successful amplification in LAMP assays, with LAB strains showing robust product formation on gel electrophoresis, whereas non-LAB controls did not amplify (Table 2), including several species of yeasts and acetic acid bacteria (data not shown).

Due to a high tendency to produce non-specific amplification, the influence of gold nanoparticles (AuNPs) was assessed. AuNPs were evaluated at volumes of 0, 1, 3, and 5 µL. At 5 µL, the negative control consistently showed no amplification, while positive samples maintained high efficiency, suggesting this as the optimal concentration (Figure 2). This improvement reinforces the role of AuNPs in reducing non-specific amplification, thus improving reaction specificity.

Multiple detection approaches were evaluated for LAMP amplicon visualisation in both white and red musts and wines. HNB produced a distinctive colour change from violet to blue in the presence of amplification, attributable to magnesium pyrophosphate formation. Neutral red provided a secondary confirmation via a pH-dependent colour shift from red to pink (Figures 3B1 and 3C1). These results were visually intensified using the HNB-enhance software tool, improving discrimination between positive and negative samples (Figures 3B2 and 3C2). In addition, amplified samples exhibited visible turbidity in the absence of any dye (Figure 3A). Gel electrophoresis confirmed successful amplification by displaying the characteristic ladder-like banding pattern (Figure 3D). While only representative data are shown in the figures, all methods were validated across both red and white wines and musts. The performance of the assay remained consistent across matrices (data not shown).

NC = Negative control

A novel semi-quantification method based on HNB colour intensity and RGB values was developed. A clear inverse relationship between the red component intensity and incubation time was observed in positive samples (Figure 4). This sigmoidal behaviour was consistent across matrices, including white and red wine and musts. No colour change was observed in negative controls, validating the specificity of the assay.

Endpoint quantification was successfully achieved using optical readings of the LAMP amplicons following an HNB-based colorimetric reaction. Standard curves generated from serial dilutions of LAB ranging from (10² to 10⁸ CFU/mL) showed a clear dose-response relationship, with higher bacterial concentrations yielding greater optical density (Figure 5). This provides a valuable and low-cost alternative to real-time fluorescence-based quantification.

Quantification of LAB using EvaGreen-based real-time LAMP showed reliable amplification kinetics. Amplification time (Tt) was inversely correlated with initial bacterial concentration (Figure 6A), and melt curve analysis demonstrated high specificity (Figure 6B). Linear regression of standard dilutions (10² to 10⁸ CFU/mL) yielded an excellent correlation (R² > 0.98), affirming the method’s quantitative potential (Figure 7).

To distinguish live cells from dead, PMA-qLAMP was employed. Without PMA, both live and heat-killed cells produced similar amplification signals. However, following PMA treatment, Tt values for dead cells were significantly delayed, or amplification was entirely inhibited, while live cells amplified normally (Figure 8). This confirms the PMA-qLAMP assay’s capability to selectively quantify viable LAB cells in complex wine and must matrices.

This study analysed four different methods, highlighting their differences in speed, accuracy, sensitivity, and cost. LAMP semi-quantification using a mobile app is fast and low-cost; however, it lacks precision. In contrast, endpoint quantification offers greater accuracy after the reaction but does not provide real-time results. Real-time LAMP enables precise and dynamic monitoring with high sensitivity, although it requires specialised equipment. Additionally, PMA-LAMP exclusively detects viable cells, making it useful for assessing microbial viability, but it necessitates extra sample preparation (see Table 3).