A lyophilization-compatible colorimetric LAMP assay for cold chain-free, extraction-free and rapid post-enrichment detection of E. coli

Abstract

Developing reliable dry-format colorimetric LAMP assays remains challenging because most existing systems depend on pH-sensitive dyes e.g. Hydroxy Naphthol Blue and Phenol Red. These dyes lose performance after lyophilization or when exposed to extraction-free sample buffers, as they require low-buffer conditions to generate accurate color transitions. Lyophilization disrupts pH stability, and common lysis additives further interfere with signal clarity. Consequently, creating a robust, field-ready colorimetric LAMP platform has been constrained. Here, we developed a Mn-PAPS Colorimetric LAMP assay, a metal ion–based colorimetric system that is inherently pH-independent and fully compatible with freeze-drying. Using Escherichia coli as a model organism, this study demonstrates the feasibility, and diagnostic performance of the lyophilized colorimetric LAMP platform. The Mn-PAPS Colorimetric LAMP platform demonstrated sensitivity in a single-digit CFU range and maintained clear visual contrast with a high analytical accuracy across 69 spiked and unspiked reference samples. Incorporation of daughter primers reduced incubation time from 120 to 45 minutes. Two lyophilized formats demonstrated room-temperature stability, supporting field deployment.

• Resolves the long-standing limitation of pH-sensitive colorimetric LAMP through a metal-ion dye system.

• Demonstrates single-digit CFU sensitivity and high analytical accuracy using a micropipette-free workflow.

• Provides stable performance under ambient conditions through lyophilization-compatible reagent formats.

Graphical abstract

Specifications table

Subject area	Biochemistry, Genetics and Molecular Biology
More specific subject area	Molecular Diagnostics
Name of your method	Mn-PAPS Colorimetric LAMP assay for E. coli detection
Name and reference of original method	•LAMP •https://doi.org/10.1016/j.mimet.2018.11.020.
Resource availability	Available upon requested

Background

E. coli is a Gram-negative, facultative anaerobic bacterium commonly found in the intestines of humans and warm-blooded animals [1]. It has long been recognized as a key indicator of fecal contamination and water quality [2]. Traditional detection methods rely on culture-based techniques, which, while reliable, are time-consuming and labor-intensive. PCR-based protocols are faster but still complicated and technically demanding [[3], [4], [5], [6]]. In 2019, we introduced a colorimetric LAMP detection platform using Xylenol Orange, a pH-sensitive dye, as a visual indicator for E. coli [7], which later supported assay development for SARS-CoV-2 [8], Mpox [9,10], and Vibrio parahaemolyticus in seafood [11,12]. Despite these advances, deploying colorimetric LAMP in resource-limited settings remains challenging. Most assays rely on liquid reagents requiring cold-chain storage, increasing logistical burden [[8], [9], [10], [11], [12], [13], [14], [15]].

Although lyophilization can stabilize reagents at ambient temperatures [16], practical implementation is limited by proprietary lyoprotectant formulations and scarce open literature on lyobead production. More critically, pH-sensitive dyes used in colorimetric LAMP require low-buffer conditions and perform poorly after freeze-drying, often producing unintended color shifts upon reconstitution [17,18]. A further complication arises when integrating extraction-free workflows. Chemical, enzymatic, or mechanical lysis methods [[19], [20], [21], [22]], commonly use surfactants and reducing agents (e.g., Triton X-100, SDS, Tween-20, TCEP) [23,24], which can alter reaction pH and disrupt color development in widely used dyes such as Phenol Red and Cresol Red [25]

To address these limitations, we developed a metal ion–based colorimetric system using 5‑bromo-2-pyridylazo-5′-N-propyl-N-sulfonopropyl aniline (5-Br-PAPS) in complex with Mn²⁺, referred to as the Mn-PAPS Colorimetric LAMP assay. This dye system is pH-independent and lyophilization-compatible. The underlying mechanism of color development is based on pyrophosphate (PPi) generated during LAMP amplification. Initially, 5-Br-PAPS forms a red-purple complex with Mn²⁺ ions. During LAMP amplification, DNA polymerase generates pyrophosphate (PPi), which binds Mn²⁺ more strongly than the dye. As Mn²⁺ is sequestered by PPi, the dye is displaced, resulting in a visible color shift from red-purple to yellow, indicating successful DNA amplification (Fig. 1A). The PPi-mediated Mn²⁺ sequestration mechanism is well documented, including the first report describing pyridylazophenol metal-sensing dyes for visual detection of DNA amplification [26] and the recent publication adopting this dye system to detect SARS-CoV-2 [27].

Fig. 1.

Principle (A), sample preparation workflow (B), and detection process (C) of the ColiDetect assay.

In contrast to the Zn²⁺/5-Br-PAPS system, the Mn²⁺/5-Br-PAPS formulation offers practical advantages for LAMP. To illustrate, Zn²⁺ complexes provide strong color contrast but are more sensitive to pH and buffer composition, complicating formulation and lyophilization, while Mn²⁺ exhibits more stable colorimetric behavior across pH conditions, enabling a fully dry, field-deployable format [27]. These considerations motivated the selection of Mn²⁺ in this study. An additional reason for selecting the 5-Br-PAPS/Mn²⁺ dye system is its color profile, which closely resembles that of Xylenol Orange (our reference dye) and phenol red, both widely used in colorimetric LAMP assay development [9,11,28,29]. Overall, this study presents a portable, dry-format colorimetric LAMP test kit, named ColiDetect, for robust visual detection, optimized for ambient storage and field deployment by using E. coli as a model pathogen.

Method details

Bacterial strain and DNA preparation

E. coli ATCC® 10536™ (Microbiologics®, MN, USA) was used as the reference strain for all experiments. The strain was first streaked onto solid agar and incubated at 37 °C for 17 hours. A single colony was subsequently transferred into 2 mL of liquid medium and cultured for an additional 16 hours under the same conditions. Genomic DNA was extracted using DNAzol® Reagent (Thermo Fisher Scientific, USA), and purity was assessed by measuring absorbance at 260/280 nm. The extracted DNA was then subjected to tenfold serial dilution to generate working stocks of 10 ng (10⁻¹), 1 ng (10⁻²), 100 pg (10⁻³), and 10 pg (10⁻⁴) per microliter (µL). Unless otherwise specified, 1 µL of each dilution was used as the template for LAMP reactions. We used total genomic DNA extracted from E. coli rather than CFU-calculated inputs because the original LAMP assay already demonstrated a limit of detection of 1 CFU in our previous work [7]. To avoid bias during optimization, the initial DNA concentration was blinded and reported only as serial dilutions (undiluted, 10⁻¹, 10⁻², 10⁻³, etc.). Although extraction efficiency may vary slightly between samples, this approach was sufficient for method development. In the final validation stage, we confirmed that the optimized assay still detects a single-digit CFU of E. coli.

Standard liquid LAMP assay for E. coli detection

The original liquid LAMP assay was performed following the protocol established in our previous work [7]. The assay targets the yaiO gene, which encodes an E. coli-specific outer membrane protein (GenBank accession no. NC_000913.3). Each 25 µL reaction contained 2 µM of inner primers (FIP and BIP), 0.2 µM of outer primers (F3 and B3) (Table 1), 4 mM MgSO₄, 0.6 M betaine, 1.4 mM dNTPs, 8 units of Bst DNA polymerase, and 1× low-reaction buffer and 0.06 mM Xylenol Orange (XO) dye. Reactions were initiated with 1 µL of DNA template and incubated at 63 °C for 120 minutes. Amplification was verified by gel electrophoresis and by visual inspection of the colorimetric indicator: a transition from purple to orange-yellow indicated a positive reaction, whereas retention of the original red-purple color signified a negative result.

Table 1.

Core primers and daughter primers. The optimized colorimetric LAMP assay in this study utilizes both core primers and daughter primers.

Name of primers	Sequence (5′−3′)
Jaroenram et al., 2019 (core primers)
yaiO2.F3	CCAGTCATAGGTGTAAGCAC
yaiO2.B3	CGTCAGGATATTAACCTGGC
yaiO2.FIP	CAGCGATGCAGGTGGTAGTT/TTTT/TAGTTGCGTATAACCAGTGC
yaiO2.BIP	GGTATAGCGGTAGCTGGTGATC/TTTT/TTACGATGATGTCGAAGTCG
yaiO2.LF	TCCGTGCGTCTGAATGAC
yaiO2.LB	GGCCAGTATAGAGTGATACGC
This study (daughter primers)
yaiO2.FIP2	ATAGCAATATGATTTTTGTTCCGCGGCTTAGCC
yaiO2.BIP2	GCCTTGCCAGGCATTTTTACGACCGGTTATCGCTAC
yaiO2.F1c-2	ATAGCAATATGA
yaiO2.B1c-2	GCCTTGCCAGGCAT

Daughter primer design and principle

In the original LAMP architecture, six primers (F3, B3, FIP, BIP, LF, LB) generate 8 initiation sites for auto-cycling amplification. To shorten the extended incubation time required by the standard LAMP protocol, an additional set of “daughter primers” was designed to anneal within the region flanked by the original FIP and BIP binding sites. These daughter primers (FIP2, BIP2, F1c-2, B1c-2) increase the total number of hybridization sites from 8 to 14, thereby accelerating strand displacement and loop formation. This nested-style primer arrangement is conceptually similar to nested PCR. Its strategy is supported by our previous work [12], which demonstrated that multiple inner primers significantly increase amplification speed and sensitivity.

To clarify primer usage in this study, two sets of primers (original and daughter) were evaluated to enhance amplification speed. The original primer set (F3, B3, FIP1, BIP1, LF, LB) was used for baseline comparison, while daughter primers (FIP2, BIP2, F1c-2 and B1c-2) were incorporated in selected experiments to assess their effect on reaction kinetics, sensitivity, and colorimetric endpoint clarity. Tables 1 and 2 summarize the primer sequences, combinations, and final concentrations used in each assay.

Table 2.

Premix Composition.

Reagent	Volume (µL)	Final Concentration
UltraPure™ DNase/RNase-Free Distilled Water	7.33	—
Core primers
100 µM yaiO2.F3	0.05	0.2 µM
100 µM yaiO2.B3	0.05	0.2 µM
100 µM yaiO2.FIP	0.25	1.0 µM
100 µM yaiO2.BIP	0.25	1.0 µM
100 µM yaiO2.LF	0.25	1.0 µM
100 µM yaiO2.LB	0.25	1.0 µM
Daughter primers
100 µM yaiO2.FIP2	0.25	1.0 µM
100 µM yaiO2.BIP2	0.25	1.0 µM
100 µM yaiO2.F1c-2	0.25	1.0 µM
00 µM yaiO2.B1c-2	0.25	1.0 µM
10× Isothermal Buffer	2.5	1X
10 mM dNTPs	3.5	1.4 mM
25% Raffinose	8	8% (w/v)
Bst Enzyme (120 U/µL)	0.4	48 U
2.5 mM 5-Bromo-PAPS	0.5	0.05 mM
1.875 mM MnCl₂	0.67	0.05 mM
Total	25	—

Real-time turbidity measurement for kinetic comparison of LAMP reactions

Real-time turbidity monitoring was performed to compare amplification kinetics between the original LAMP assay [7] and the daughter-primer LAMP assay (this study). Reactions were run on a Loopamp Realtime Turbidimeter (LA-500; Eiken Chemical Co., Japan), which continuously measures turbidity generated by magnesium pyrophosphate, a by-product of LAMP amplification. Each reaction was prepared using the same liquid LAMP formulation described above, with or without daughter primers, and incubated at 63 °C. The instrument automatically recorded turbidity values at fixed intervals and displayed amplification curves in real time. This approach allowed direct visualization of reaction kinetics rather than relying solely on endpoint colorimetric readouts, enabling assessment of how daughter primers influenced amplification speed and onset time.

Endpoint colorimetric enhancement with daughter primers

Colorimetric LAMP reactions of our original liquid LAMP with the addition of daughter primers were incubated at 63 °C in a heating device, and tubes were briefly removed at 15-minute intervals (15, 30, 45, and 60 minutes) to document color transitions before being returned to the heating device. Incubation times for endpoint colorimetric evaluation were selected based on the real-time turbidity profiles obtained from the kinetic comparison experiments.

Development and optimization of lyophilized LAMP premix

To establish a dry-format LAMP system, the original liquid assay was reformulated by replacing the Xylenol Orange indicator with the metal ion–based 5-Bromo-PAPS/MnCl₂ dye system. Preliminary optimization experiments were performed to determine suitable concentrations of 5-Bromo-PAPS, Mn²⁺, and incubation parameters that would yield a clear visual distinction between positive and negative reactions after amplification. Raffinose (Sigma) at 8% w/v was selected as the lyoprotectant based on prior optimization work from our laboratory involving LAMP assays for Mycobacterium tuberculosis (data not shown).

Lyophilization was carried out using a Genesis Pilot Freeze Dryer (SP Industries, Inc., USA) equipped with programmable shelf temperature and vacuum control. The freeze-drying cycle consisted of three stages. Freezing: samples were held at −45 °C for 120 minutes to ensure complete solidification. Primary drying: the shelf temperature was maintained at −45 °C for 240 minutes, then gradually increased to −40 °C at 0.5 °C/min and held for 720 minutes under 200 mTorr to sublimate unbound water. Secondary drying: the temperature was raised to +25 °C at 0.5 °C/min and maintained for 720 minutes at 200 mTorr to remove residual moisture and stabilize the reagents. Each temperature transition included a 60-minute hold to prevent collapse of the lyophilized matrix. Following lyophilization, tubes were transferred to a Vigor glovebox (Vigor Technologies Inc., Texas, USA) to ensure ultra-low-moisture conditions during sealing. The glovebox maintains a default moisture level of 0.01 ppm. When lyophilized tubes are first introduced, the moisture level temporarily increases (0.5–1 ppm depending on batch), and the tubes are only sealed once the sensor readings return to 0.01 ppm. Although we did not directly measure moisture inside individual tubes, this controlled environment ensures that prototype kits are sealed under extremely dry conditions. After 24 h, lyophilized reagent pellets were sealed and stored at −20 °C till use. To use, they were reconstituted with distilled water and tested using DNA templates of varying concentrations. Performance metrics, including sensitivity, clarity of the colorimetric transition, and the potential for time-dependent nonspecific amplification caused by primer self-polymerization, were compared with those of the liquid-phase LAMP assay.

Specificity testing of the lyophilized LAMP

Assay specificity was assessed using genomic DNA from a diverse panel of bacterial species. DNA samples (100 ng each) were obtained from the repository of the Bioengineering and Sensing Technology Research Team (IBST), National Center for Genetic Engineering and Biotechnology (BIOTEC), Thailand. The panel included Vibrio parahaemolyticus, Vibrio cholerae, Bacillus cereus (BCC 6386), Listeria monocytogenes (ATCC 19115), Listeria innocua, Pseudomonas aeruginosa (ATCC 27,853), Staphylococcus aureus, Enterobacter spp., Shigella sonnei, and Acinetobacter baumannii (ATCC 19,606). A no-template reaction served as the negative control, while E. coli DNA was used as the positive control.

Preparation of samples for CFU determination and diagnostic evaluation

To assess the analytical performance of the colorimetric LAMP assay, E. coli samples with defined colony-forming unit (CFU) inputs were prepared. The objectives were to (i) quantify viable cell numbers through serial dilution and plate counting, (ii) generate standardized CFU concentrations for spiking experiments, and (iii) compare assay sensitivity using pre-enriched samples with and without heat treatment. A culture of E. coli ATCC® 10,536™ was initiated by inoculating a single colony into 5 mL of LB broth, followed by incubation at 37 °C for 16 hours. Serial dilutions were prepared using six 1.5 mL microcentrifuge tubes, each containing 900 µL of LB broth. A 100 µL aliquot of the overnight culture was added to the first tube (10⁻¹), mixed thoroughly, and subsequently diluted stepwise to 10⁻⁶. From dilutions 10⁻³ through 10⁻⁶, 100 µL of each dilution was plated onto LB agar and spread evenly using a sterile spreader. Plates were incubated at 37 °C for 16 hours, and colonies were counted on plates containing 10–300 CFU. Representative colony counts were 110 CFU for the 10⁻³ dilution, 13 CFU for 10⁻⁴, and 1 CFU for 10⁻⁵, confirming expected dilution performance. Based on these values, working dilutions were prepared from the 10⁻³ stock to yield approximate reaction inputs of 11, 1.1, and 0.11 CFU per tube. These standardized CFU levels were used in all subsequent diagnostic evaluations.

Evaluation of heat-treated vs. untreated samples

Sample matrices included autoclaved water (50 µl), sterilized milk (50 µl), and cooked chicken meat (0.5 cm³). Each matrix was spiked with 50 µl of E. coli at four concentrations (0.11, 1.1, 11, and 110 CFU), followed by pre-enrichment in 900 µl LB broth at 37 °C for 16 h with shaking at 220 rpm. To assess whether thermal lysis enhances DNA release and detection sensitivity, samples were then split into two treatment conditions: heat-treated (boiled for 10 min) and untreated (room temperature). One microliter of each sample was added to LAMP reaction tubes containing lyophilized colorimetric master mix reconstituted with 24 µL of distilled water. After 45 min incubation the results were compared to that of the freshly prepared LAMP reactions.

Blind testing of the prototype kit using food samples

Blind diagnostic evaluation was performed using 69 food samples, consisting of 34 liquid matrices (bottled water and sterilized milk) (50 µl each) and 35 solid matrices (cooked chicken meat, ∼0.5 cm³). Positive samples were spiked with 50 µl of E. coli at 1.1, 11, or 110 CFU, while negative controls received 1 µL of distilled water. All samples underwent pre-enrichment and heat treatment as described above. For testing, lyophilized LAMP reagents were reconstituted with 25 µL of distilled water. A sterile 1 µL inoculating loop was used to transfer each sample into a reaction tube. The loop-based transfer method was selected to simplify field operation, as it introduces a small but consistent volume of sample without requiring micropipettes or disposable tips. After sealing, tubes were incubated at 63 °C, and photographs were taken every 15 minutes for up to 60 minutes. Final colorimetric outcomes were interpreted by comparing tube colors with predefined standards. This simplified workflow was designed to enhance usability for non-specialist operators and to support deployment in resource-limited environments. In this workflow, the limit of detection refers to the initial viable-cell input added prior to enrichment, not to the concentration of the enriched broth or to the 1 µL loop volume used for transfer. After 16 h enrichment, the inoculating loop serves only as a consistent transfer tool and does not define analytical sensitivity. Thus, the assay detects an initial input of ∼1 CFU per sample, consistent with enrichment-based food microbiology workflows.

Design, assembly, and operational workflow of the colidetect kit

Following completion of assay validation, we assembled ColiDetect, a ready-to-use diagnostic kit designed for field deployment. The kit is available in two configurations: All-in-One, which contains lyophilized primers, enzymes, buffer, and dye in a single tube, and Dye-free, which requires the addition of a reconstitution buffer (Supplementary Method 1: Test Kit Specification). Each kit provides materials for 48 tests and includes positive and negative controls, enrichment media, and essential accessories. A detailed list of components, along with representative images of the lyophilized LAMP cakes, is provided in Table S1.

The operational workflow, summarized in Fig. 1B and 1C, consists of three main steps. Step 1: Sample Preparation and Pre-enrichment. Basic materials such as gloves, forceps, and a heating block are prepared. Solid food samples are cut into small pieces and transferred into enrichment tubes containing 950 µL of media. For liquid samples, approximately 50 µL is added directly. Tubes are incubated at 37 °C for 16 hours with shaking. An optional heat-treatment step (100 °C for 10 minutes) is recommended to enhance DNA release and improve assay consistency Step 2: Testing Procedure. Lyophilized reagents are reconstituted with 25 µL of distilled water (All-in-One format) or reconstitution buffer (Dye-free format). A sterile inoculating loop is used to transfer the Positive Control into the first tube, test samples into subsequent tubes (using a new loop for each sample), and the Negative Control into the final tube. Tubes are sealed and briefly centrifuged or flicked to ensure the liquid settles at the bottom. Reactions are incubated at 63 °C for 45 minutes. Step 3: Result Interpretation. Colorimetric outcomes are evaluated visually: a yellow color indicates a positive reaction, while a reddish-pink color indicates a negative result. The Positive Control must turn yellow, and the Negative Control must remain reddish-pink; any deviation renders the run invalid and requires retesting. Troubleshooting guidance is provided in Supplementary Method 2.

Shelf-life evaluation

The long-term stability of the ColiDetect kit was assessed by storing lyophilized reagents at 37 °C for defined time intervals and periodically testing their performance. Sensitivity was evaluated using E. coli samples with known CFU inputs prepared earlier in the study. To estimate performance stability under different storage conditions, projected shelf-life values at 25 °C, 4 °C, and −20 °C were calculated using the Q10 Accelerated Ageing model [30,31] (Supplementary Method 3). This model predicts degradation rates based on temperature-dependent reaction kinetics. The resulting data were used to determine practical storage limits and to identify conditions under which the kit maintains reliable diagnostic performance.

Reproducibility and inter-observer agreement study

Three types of reliability were evaluated: intra-assay reproducibility, inter-assay reproducibility, and inter-observer agreement. These assessments were performed using two concentrations of E. coli: 1 CFU per reaction (LOD) and 10⁸ CFU per reaction, along with a distilled-water negative control. Each condition was tested in triplicate. As for the intra-assay assessment, nine reactions (3 concentrations × 3 replicates) were processed simultaneously under identical conditions by a single operator. This provided a baseline measure of consistency within a single run. The inter-assay evaluation followed the same layout but was repeated on three separate days using the same batch of lyophilized reagents, yielding 27 reactions in total. Colorimetric outcomes were recorded at the 45-minute endpoint, and overall stability was expressed as the percentage of matching results across all runs. Data from Day 1 also served as the intra-assay dataset. Finally, inter-observer agreement was examined using the same 27 tubes. Three independent evaluators, blinded to sample identities, scored the final tube colors. Concordance among their interpretations was calculated as the percentage of matching classifications, providing insight into how reliably the assay can be interpreted by different users.

Method validation

Real-time kinetic comparison of LAMP reactions with and without daughter primers

Real-time turbidity monitoring revealed clear differences in amplification kinetics between the original LAMP assay and the daughter-primer LAMP assay. The original LAMP reaction, which uses only the core primer set, began to show a detectable turbidity increase at approximately 45 minutes with cumulative turbidity (Abs at 650 nm) peaking just below 0.5 units. (Fig. 2A). In contrast, the daughter-primer LAMP reaction exhibited accelerated kinetics, with a positive turbidity signal appearing at ∼30 minutes, and cumulative turbidity (Abs at 650 nm) reached ∼1.0 at reaction completion—nearly twice that of the conventional LAMP in panel A—indicating both faster and stronger amplification (Fig. 2B). These results confirm that increasing the number of hybridization sites substantially enhances auto-cycling efficiency and shortens the time required for amplification onset. Late-stage non-specific amplification was observed in negative controls after ∼90 minutes when daughter primers were used, indicating that prolonged incubation increases the risk of false-positive signals. This observation guided the selection of an appropriate endpoint incubation time for colorimetric detection in the next experiment.

Fig. 2.

Effect of daughter primers on accelerating LAMP reaction kinetics and endpoint color development using 100 pg E. coli DNA (equivalent to 2 × 10⁴ CFU; P) and a template-free negative control (N). Please refer to Supplementary Method 4 for the genomic DNA–to–CFU conversion. (A) Real-time turbidity monitoring of the original LAMP assay (without daughter primers). (B) Real-time turbidity monitoring of LAMP reactions containing daughter primers. (C) Endpoint colorimetric LAMP results with the addition of daugther primers.

Endpoint colorimetric enhancement with daughter primers

To establish an appropriate endpoint incubation time for our colorimetric LAMP assay, it is essential to first consider the reaction kinetics governing dye conversion and polymerase performance. Endpoint colorimetric LAMP relies on visible dye transition rather than real-time signal monitoring, meaning that incubation time must be long enough to accumulate sufficient amplification products while still preventing non-specific reactions. This balance is particularly critical when detecting low DNA inputs.

In our previous study [7], the original liquid LAMP assay—based on a pH-sensitive dye system—required up to 120 minutes to detect single-digit CFU inputs (Fig. S1). Such extended incubation reflects the kinetics of the XO dye, which demands substantial pyrophosphate (PPi) accumulation to drive a large pH shift (from ∼8.5 to ≤6.7) for visible color change. Although amplification was detectable by gel electrophoresis within 60–75 minutes, XO responds slowly, especially when template concentration is very low (i.e. <100 CFU). Theoretically, this lengthy incubation can be shortened by using advanced LAMP polymerases. Several fast enzymes—such as Bst 3.0 (NEB) and Lyo-Ready Bst DNA Polymerase (Invitrogen™)—can complete amplification within 10–30 minutes, but they also produce non-specific amplification when reactions extend beyond ∼30 minutes [32,33]. Because endpoint colorimetric LAMP requires longer incubation to detect low DNA inputs while maintaining strict positive/negative discrimination, these rapid enzymes are not chosen here. The conventional Bst 2.0 WarmStart polymerase (NEB), which amplifies more slowly (typically requiring 45–60 minutes) [34] but provides higher specificity, was therefore selected for endpoint detection in this study.

With daughter primers, the LAMP assay showed a clear red-to-yellow transition after 45 minutes of incubation in the XO-based LAMP reactions (Fig. 2C), which is more than twice as fast as our conventional LAMP assay (Fig. S1). This phenomenon aligns with the stronger amplification and higher cumulative turbidity/PPi observed in the real-time turbidity assessment (Fig. 2B). However, to maximize color development while minimizing false-positive outcomes—which may appear after ∼90 minutes of incubation based on the results in Fig. 2B—a 45-to-60-minute incubation was selected for subsequent experiments.

Optimization of lyophilized LAMP conditions

• •Optimal Dye Concentration

To determine the most effective formulation for the metal-ion dye system, various concentrations of 5-Bromo-PAPS and MnCl₂ were evaluated. As shown in Fig. 3, combinations containing 50 or 75 µM 5-Bromo-PAPS paired with 50 µM MnCl₂ produced the strongest contrast between positive reactions (10⁻³ dilution; 100 pg) and the negative control (N) after 45 minutes at 63 °C (highlighted in red boxes). Extending incubation beyond 60 minutes reduced color contrast and made interpretation less straightforward. Based on these observations, 50 µM 5-Bromo-PAPS and 50 µM MnCl₂ were selected as the optimal concentrations for all subsequent experiments.

• •Lyophilization Conditions and Performance Comparison

Fig. 3.

Optimization of 5-Bromo-PAPS and MnCl₂ concentrations and incubation time for colorimetric LAMP. The positive and negative reactions are denoted by 10^–3 dilution (100 pg DNA/reaction) and N (DNA-free), respectively. Slight background color differences in negative tubes after extended incubation (>60 min) are consistent with the chemistry of metal-ion-based colorimetric LAMP, arising from minor shifts in the Mn²⁺–PPi–5-Br-PAPS equilibrium, small batch-to-batch reagent variations. Prolonged incubation may also promote late non-specific amplification or primer-dimer formation (Refer to Fig. 2B further discussion).

The lyophilized LAMP premix formed a uniform purple-pink cake within each reaction tube (Fig. 4A). Upon reconstitution with distilled water, the dried reagents dissolved rapidly and returned to their characteristic reddish-purple color (data not shown). Performance testing showed that the lyophilized formulation remained stable even when incubation exceeded 120 minutes (Fig. 4B). No false-positive reactions were observed, and the colorimetric transition remained sharp. We hypothesize that lyophilization stabilizes the LAMP mechanism and actively reduces non-specific amplification. This hypothesis is supported by our real-time turbidity measurements; while the liquid LAMP formulation containing betaine (without cryoprotectants) exhibited non-specific amplification after 90 min (Fig. 2B), the lyophilized assay showed no non-specific amplification even after 120 min of incubation. This observed stabilization aligns with recent findings indicating that the lyophilized version of a LAMP assay demonstrated fewer false positives than its identical liquid counterpart [35]. This improvement may be attributed to the protective effects of the added cryoprotectant, as well as the physical immobilization of the freeze-dried master mix. Because the components remain completely dry until sample hydration at the exact onset of incubation, the mix effectively functions as a pseudo-hot start mechanism, narrowing the time window in which low-temperature, non-specific primer interactions can occur. Further studies are required to fully validate this concept. This behavior closely matched that of the liquid-phase LAMP assay [7], indicating that the freeze-drying process did not compromise reaction fidelity. However, to balance between color development and avoid non-specific dimerization that might occur in excessive incubation, 45 - 60 min incubation is recommended. The final reagent composition (pre-freeze volume: 25 µL) is summarized in Table 2.

Fig. 4.

Lyophilization results. (A) Example of appearance of lyophilized, dye-included LAMP reagents in its dry “cake” form after moisture removal (referred to as All-in-One format). (B) Sensitivity representative of dye-included lyophilized liquid LAMP reagent with the addition of daughter primers (adapted from Jaroenram et al., 2019). After reconstitution with water and incubation, only reactions with sufficient E. coli DNA show a clear color change. Note: Samples 10⁻³, 10⁻⁴, 10⁻⁵ dilutions and N correspond to 100 pg, 10 pg and 1 pg of E. coli DNA, and template-free reaction, respectively.

Validation of specificity testing

The assay showed specificity for E. coli. None of the non-target bacterial species turns yellow (Fig. 5). These results confirm that the additional daughter primer set and dye chemistry do not cross-react with closely related or environmentally relevant organisms. The current specificity test did not include certain enteric relatives such as Klebsiella pneumoniae and Citrobacter freundii because they were not available at the time this study was being conducted. However, we performed BLAST analyses using both core and daughter primers against the complete genomes of K. pneumoniae subsp. SB3432 (GenBank: FO203501.1) and C. freundii B38 (GenBank: CP016762.1). No significant primer binding was observed (Fig. S3), suggesting low cross-reactivity risk. Nonetheless, we acknowledge that wet-lab confirmation remains necessary, and these species should be incorporated into future specificity testing.

Fig. 5.

Specificity of the developed test kit using DNA samples from various bacterial species. (A) Colorimetric readout of the lyophilized LAMP reactions. (B) Corresponding gel electrophoresis results confirming amplification outcomes. As expected, tubes showing reddish-pink coloration do not contain LAMP amplification products, whereas the yellow tube corresponds to a true positive reaction.

Validation of heat treatment effects on lyophilized LAMP performance and sensitivity

Both heat-treated and non-heat-treated samples yielded comparable results, with a consistent detection sensitivity of a single-digit CFU (approximately 1.1 CFU per reaction) (Fig. 6A). This detection limit aligns with that of the standard liquid LAMP formulation previously reported [7], indicating no loss of sensitivity following lyophilization. These findings also confirm that both treatment conditions are compatible with diverse sample matrices and preserve assay sensitivity. It aligns with previous studies demonstrating that extraction-free LAMP assays can reliably detect target DNA without purification steps [36]. However, we encourage readers to use the boiling method to ensure that E. coli cells are completely disrupted to release DNA [37] and that any viable bacteria are fully inactivated. The use of LB broth as an enrichment medium did not inhibit the LAMP reaction, consistent with reports that LAMP assays tolerate residual culture media provided the input volume remains below inhibitory thresholds [38]. This compatibility simplifies sample preparation and supports broader deployment in food safety testing.

Fig. 6.

Comparison of lyophilized LAMP sensitivity and sample preparation methods (heat-treated vs. non-heat-treated). Water, milk, and cooked chicken samples were spiked with E. coli at concentrations of (A) 0.11 and 1.11 CFU, and (B) 11 and 110 CFU. All samples underwent pre-enrichment, followed by either heat treatment or no treatment, and were tested using the lyophilized LAMP kit. Colorimetric results were recorded at 45 min of post-incubation. Across the lowest dilution series, reactions receiving at least one viable cell consistently produced a positive colorimetric signal, demonstrating sensitivity within the single-digit CFU range. Minor background color variation in negative controls is inherent to metal-ion colorimetric LAMP systems, arising from small shifts in the Mn²⁺–PPi–5-Br-PAPS balance and batch-to-batch reagent differences, as well as lighting conditions during image capture.

Validation of assay performance by food samples

To evaluate analytical performance under conditions resembling real use, blind testing was conducted on 69 food samples. Of these, 45 were spiked with E. coli (positive group) and 24 were unspiked (negative group). When results were interpreted at the 45-minute endpoint, the assay achieved 100% sensitivity, 100% specificity, and 100% accuracy (Table 3). These findings demonstrate that the ColiDetect workflow performs reliably across diverse food matrices and supports accurate visual interpretation without specialized equipment. Although the assay is DNA-based, its intended use is after enrichment, which aligns the workflow with viability-based detection rather than direct DNA detection. In food microbiology, culture-based enumeration remains the regulatory gold standard for determining viable E. coli contamination levels. Enrichment ensures that the assay reflects viable cell presence and minimizes false-positive bias from dead-cell DNA, a known limitation of direct qPCR workflows. For this reason, culture-based results were included to contextualize the practical application of the assay within standard food testing workflows.

Table 3.

Diagnostic Performance Summary of ColiDetect from Blind Sample Testing.

Default statuses of samples	ColiDetect (this study)
	Positive	Negative
Positive (with E. coli)	45 (TP)	0 (FN)
Negative (without E. coli)	0 (FP)	24 (TN)
Summary of diagnostic performance	100 (sensitivity) (95% CI: 92.1–100%)	100 (specificity) (95% CI: 85.8–100%)
	100 (accuracy) (95% CI: 94.8–100%)

Sensitivity (the ability to identify positive samples correctly) = [TP/(TP+FN)]×100, Specificity (the ability to identify positive samples correctly) = [TN/(TN+FP)] ×100.

Accuracy (the ability to identify both positive samples and negative sample correctly) = [(TP+TN)/(TP+TN+FN+FP)] ×100.

TP: true positive, FP: false positive, FN: false negative, TN: true negative.

Note: Diagnostic metrics were derived from 69 validated samples. A statistical power-based sample size calculation (e.g., Buderer’s formula) will be applied in subsequent work.

Validation of shelf-life of the colidetect test kit

The stability of the ColiDetect assay was next examined under accelerated ageing conditions. For the All‑in‑One formulation (Lyo A), the lyophilized reagents remained fully functional for 26 days at 37 °C, full detection limit within the single-digit CFU range. Based on the Q10 ageing model, this corresponds to an estimated shelf life of 2 months at 25 °C, 8 months and 21 days at 4 °C, and approximately 3 years and 10 months at −20 °C. Beyond this period, performance declined: after 39 days at 37 °C, sensitivity decreased to 10 CFU, and by day 46 the lyophilized matrix no longer dissolved cleanly, forming a glue‑like residue (data not shown) accompanied by complete loss of amplification. The Dye‑free formulation (Lyo B) showed a similar stability profile during the first 26 days at 37 °C; however, by day 39, color development weakened, making interpretation difficult. Follow‑up testing revealed that this decline was caused by degradation of the reconstitution buffer rather than instability of the lyophilized pellet. When freshly prepared buffer was used, full sensitivity (the single-digit CFU range) was restored even after 60 days at 37 °C (data not shown). Shelf‑life projections for Lyo B were therefore longer: 4.5 months at 25 °C, 19 months at 4 °C, and approximately 8.5 years (103 months) at −20 °C. A summary of these findings is provided in Table 4, with complete data sets in Table S2.

Table 4.

Shelf-Life Performance of ColiDetect Test Kit using accelerated ageing calculator [30]. Testing concluded at Day 60 due to depletion of test kit stock (See Supplementary Result 2 for experimental images).

Format	LoD (Initial)	Stability at 25 °C	Stability at 4 °C	Stability at −20 °C
All-in-One	Single-digit CFU range	2 months	8 months	3 years and 10 months
Dye-Free	Single-digit CFU range	4 months and 2 weeks	1 year and 7 months	8 years and 6 months

Notes: LoD = Limit of Detection.

The performance degradation observed in the All‑in‑One format after extended storage may be influenced by moisture‑related factors, including incomplete moisture removal during lyophilization and moisture ingress during post‑drying handling or incubation. This interpretation is supported by the observation that, by day 46 at 37 °C, the lyophilized cake in the All‑in‑One tubes began to partially melt (data not shown), coinciding with complete loss of amplification. Such physical softening is consistent with increased residual moisture, which accelerates enzyme degradation and destabilizes lyophilized matrices by lowering the glass transition temperature (Tg) and increasing molecular mobility [39]. This behavior aligns with recent findings showing that excipient composition and residual water strongly influence Tg′ and amorphous‑phase stability in lyophilized systems [40]. In addition to general moisture ingress, the hygroscopic nature of MnCl₂ may further contribute to localized increases in water activity within the lyophilized matrix. MnCl₂ is well documented to absorb moisture readily and form stable hydrates [41,42], and this property may amplify moisture-driven destabilization in dye-containing pellets even when sealed under ultra-low-humidity conditions. Because the only compositional difference between the two formats is the inclusion of the 5-Br-PAPS/Mn²⁺ dye system in the All-in-One pellets, it is plausible that MnCl₂ and its associated buffer components may alter matrix behavior—for example, by modifying solute–water interactions or affecting moisture distribution—thereby increasing susceptibility to humidity-induced degradation. Notably, this melting phenomenon was not observed in the Dye‑free format under identical conditions. Although no studies have directly evaluated the impact of metal‑ion dyes on Tg in lyophilized diagnostic reagents, recent work demonstrates that added solutes can significantly alter matrix behavior and moisture sensitivity [43]. Thus, the dye‑containing formulation may be inherently more susceptible to moisture‑driven destabilization, although this remains a hypothesis. Future formulation optimization, including controlled‑humidity handling, direct moisture quantification, and excipient screening, will be required to determine whether the dye contributes mechanistically to the increased moisture sensitivity observed in the All‑in‑One format.

As for the performance decline of the Dye‑free format, the loss of sensitivity was likely driven by degradation of the reconstitution buffer rather than instability of the lyophilized reagent. This was evident on day 46 of incubation: when freshly prepared buffer was used, assay sensitivity immediately returned to 1 CFU and remained stable through day 60. This indicates that the lyophilized reagent itself remained functional and that buffer integrity was the limiting factor. Preparing the buffer fresh from refrigerated stock solutions may therefore help preserve assay performance during extended storage. On the whole, ColiDetect is available in two formats: an All-in-One format, which is fully premixed and requires only distilled water but exhibits shorter shelf-life due to moisture sensitivity as discussed above, and a Dye-free format, which offers extended stability because the reconstitution buffer and dye are added fresh at use.

Validation of reproducibility and inter-observer agreement study

To ensure that ColiDetect performs reliably, we evaluated its reproducibility across multiple dimensions. Both intra-assay and inter-assay tests showed complete consistency across all replicates at the limit of detection within the single-digit CFU range and at high bacterial loads (10⁸ CFU), even when experiments were repeated on three independent days (Fig. S2, Table S3). The visual endpoint also proved highly objective: three blinded evaluators reached identical interpretations for all tubes. While these results support the robustness of the visual readout, future work should incorporate quantitative absorbance-based colorimetry to better distinguish true performance decline from procedural variation and to establish standardized thresholds for long-term stability assessments.

Comparative analysis with other commercial kits

In comparative analysis, commercial kits such as Compact Dry EC, Bactaslyde, and Aquagenx offer user-friendly formats and visual indicators but are limited by long incubation times (up to 24 h), reduced specificity, and incompatibility with lyophilization. For example, Compact Dry EC and Bactaslyde rely on bacterial growth on agar plates or nutrient-coated strips, with reported specificity ranging from 66% to 95.2%, while Aquagenx offers extended shelf-life but lower overall accuracy (87.5%). In contrast, ColiDetect achieved 100% analytical accuracy within the tested set of 69 diverse samples (Table 3) within the single-digit CFU range (Fig. 6, Table 4). ColiDetect also outperforms several recently reported E. coli colorimetric platforms. DNAzyme-LAMP assays combining colorimetric and electrochemical detection report LoDs between 6.3 and 63 CFU [44]; lateral flow chromatographic LAMP systems range from 5.7 to 5.7 × 10² CFU [45]; and aptamer-based assays using gold nanoparticles with peroxidase-like amplification yield LoDs around 5 × 10³ CFU [46]. These platforms require post-amplification steps such as hybridization, signal development, or specialized readout devices. In contrast, ColiDetect enables direct visual interpretation without auxiliary workflows, reducing contamination risk and simplifying field deployment.

Overall, this study overcomes the major limitations of conventional colorimetric LAMP platforms that rely on pH-sensitive dyes, which require low-buffer conditions and are highly susceptible to pH fluctuations. Such dyes are incompatible with lyophilization and prone to unreliable color shifts, particularly when exposed to acidic or alkaline extraction reagents such as guanidine thiocyanate or sodium hydroxide [18,[32], [33], [34]]. The optimized workflow—comprising sample collection (∼1 minute), pre-enrichment (16 hours), recommended heat treatment (10 minutes), and LAMP amplification (45–60 minutes)—provides a practical balance between sensitivity and operational simplicity. Notably, heat-free samples following pre-enrichment remained compatible with the ColiDetect assay (Fig. 6).

Building on these improvements, the ColiDetect prototype delivers several advances over traditional LAMP-based diagnostics. (1) Process innovation: The use of a metal-ion indicator (5-Br-PAPS/Mn²⁺) overcomes the pH-sensitivity issues that limit conventional colorimetric LAMP assays and enables reliable lyophilization and extraction-free workflows. (2) Operational simplicity: Transitioning from micropipette-dependent workflows to inoculating-loop transfer reduces technical barriers and supports deployment in low-resource settings. (3) Stability engineering: The introduction of the Dye-free format extends shelf life to 17 months at 4 °C, addressing a major bottleneck in existing liquid colorimetric LAMP diagnostics. Together, these improvements position ColiDetect as a practical platform for rapid E. coli detection, with performance characteristics that exceed those of many existing commercial and research-grade systems. Although the assay requires a 16-hour enrichment step, enrichment is already standard practice in food microbiology to ensure detection of viable E. coli. The added value of our method lies in the post-enrichment phase: culture-based confirmation typically requires 18–48 h, as described in ISO 16,649-2 [47] and FDA BAM Chapter 4 [48], whereas the developed LAMP assay provides results within 45 min post-enrichment. The extraction-free, lyophilization-compatible format further enables rapid, equipment-free post-enrichment deployment, offering a practical alternative to lengthy culture workflows.

Table 5AU: Table 5 were not cited in the text. Please check the suggested citations and correct if required.

Table 5.

Comparative analysis of E. coli detection technologies.

Comparison Criteria	ColiDetect (This Study)	Compact Dry EC (E. coli & Coliform)	Bactaslyde E. coli & TBC	Aquagenx Define Water Test Kit
Format & Functionality	Dry reagent in PCR tube; ready-to-use; simple visual readout; no technical expertise required	Pre-prepared agar plate; results within 24 h	Nutrient-coated test strip for bacterial growth; visual readout	Powdered growth medium added to water sample; visual indicator
Equipment & Facility Requirements	Requires heating block; no complex equipment	No need for large lab or complex equipment	No need for large lab or complex equipment	No need for large lab or complex equipment
Complexity	Simple setup	Simple setup	Simple setup	Simple setup
Performance (Sensitivity, Specificity, Accuracy)	100% across all categories LoD = within a single-digit CFU	LoD = 1 CFU	Quantitative range: 10²–10⁷ CFU Sensitivity: 94.4%–97% Specificity: 66%–95.2%	LoD = 1 CFU Accuracy: 87.5%
Cost per Test	∼$ 6.4	∼$ 3	∼$ 4.6	∼$ 12.4
Shelf-Life	All-in-One: 2 months at 25 °C, 8 months at 4 °C Dye-free: 4.5 months at 25 °C, 19 months at 4 °C	1 month at room temperature	8 months from manufacturing date	3–5 years at room temperature

Limitations

• •Moisture sensitivity in the All-in-One format: Extended incubation at 37 °C beyond 46 days caused partial melting of the lyophilized cake and loss of amplification, likely due to residual moisture or moisture ingress. The specific contribution of the metal ion dye system to this effect remains unresolved. In applications where long-term stability is critical, the Dye-free format is recommended, as it remained physically stable throughout the entire shelf-life study.
• •Buffer instability in the Dye-free format: Sensitivity loss was linked to degradation of the reconstitution buffer rather than the lyophilized reagent, indicating that buffer freshness is essential for long-term performance.
• •Endpoint-only detection: The assay relies on visual endpoint interpretation, which limits quantitative assessment and may reduce sensitivity to early performance decline.
• •Environmental robustness not fully assessed: Regarding the stability claims, we have refined our specifications to reflect controlled room temperature (25 °C) storage. While our accelerated aging data at 37 °C provided a baseline for shelf-life estimation, we acknowledge that tropical ambient temperatures during transport can exceed these conditions. Consequently, we recommend that ColiDetect be transported and stored under refrigerated conditions (4 °C or −20 °C) to ensure maximum enzyme stability and extend shelf-life. The primary advantage of this lyophilized format remains its robustness during on-site application, where it can be handled at ambient temperatures without immediate degradation. Future stress testing at 50–60 °C is planned to further validate its resilience against extreme logistics environments in tropical regions.
• •Absence of parallel testing with a reference method (qPCR): The present evaluation relied on spiked and unspiked reference samples without comparison to an independent gold-standard method such as qPCR. While contrived-sample validation is widely used for early-stage assay development [49,50], the lack of parallel qPCR testing means that false-positive and false-negative rates under natural contamination conditions remain unverified. A systematic follow-up study using naturally contaminated food samples and multiple reference methods including qPCR conventional PCR, culture-based enumeration, and LAMP assay [51] should be planned to address this limitation.

Ethics statements

None.

Declaration of generative AI and AI-assisted technologies in the manuscript preparation process

During the preparation of this work, the author(s) used ChatGPT to check grammar and readability of the manuscript. After using this tool, the author(s) reviewed and edited the content as needed and took full responsibility for the content of the published article.

CRediT authorship contribution statement

Wansadaj Jaroenram: Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing. Yada Ongchanchai: Methodology. Prachumporn Nounurai: Writing – review & editing. Supattana Janta: Resources. Wansika Kiatpathomchai: Resources.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

Data will be made available on request.