Pneumonia is a major global disease and the leading infectious cause of death in children under five disease according to the World Health Organisation (WHO) (Pneumonia in children, 2022). In 2019, the WHO and the United Nations International Children’s Emergency Fund (UNICEF) estimated that 84 children die from pneumonia every hour, mostly in sub-Saharan Africa and South Asia (Pneumonia in children, 2022; Pneumonia in children: What you need to know | UNICEF, 2024). The actual burden is likely underestimated due to unreported cases.

Pneumonia is caused by bacteria, viruses, and fungi. According to the WHO, the main pathogens are Streptococcus pneumoniae, Haemophilus influenzae, respiratory syncytial virus (RSV), and Pneumocystis jirovecii (Pneumonia in children, 2022). While pathogen prevalence varies depending on factors such as geography and vaccination coverage, several remain consistently common worldwide, including S. pneumoniae, RSV, rhinovirus (HRV), human metapneumovirus (hMPV), influenza viruses, Moraxella catarrhalis, and Mycoplasmoides pneumoniae (Bénet et al., 2017; Pneumonia Etiology Research for Child Health (PERCH) Study Group, 2019). Pneumonia in children: What you need to know | UNICEF (2024) study of pneumonia-related death in South Asia and sub-Saharan Africa found S. pneumoniae, Klebsiella pneumoniae, and H. influenzae as the primary pathogens (Mahtab et al., 2024).

The timely and accurate identification of the causative pathogen in childhood pneumonia is essential to guide appropriate treatment and reduce associated mortality. Each hour of treatment delay increases the risk of death (Jones et al., 2020). While bacterial culture is the gold standard, it takes 24–72 h. PCR-based methods are faster but costly and require complex equipment, limiting their use in high-burden regions. Thus, alternative diagnostic methods must be rapid, reliable, and accessible where necessary.

Molecular techniques like loop-mediated isothermal amplification (LAMP) offer promising alternatives to conventional methods and have been widely used in recent years for pathogen detection (Mori and Notomi, 2009; Atceken et al., 2023). First described in 2000 (Notomi, 2000). LAMP can detect specific DNA and RNA sequences at a constant temperature, making it simpler and more accessible than PCR, as it requires no complex equipment. Results can be obtained in less than an hour, with performance comparable to, or even better than, PCR (Foo et al., 2020; Inaba et al., 2021).

A key advantage of this technology is the simplicity of distinguishing positive from negative reactions. This can be done by measuring turbidity due to the large amount of products (Yuan et al., 2019), or by a simple electrophoresis to observe a characteristic amplification banding pattern (Carvajal-Gamez et al., 2024).

Fluorescent and colorimetric dyes are widely used in LAMP reactions. During amplification, pH indicators, including phenol red, neutral red, and cresol red, undergo a colour change during amplification (Huang et al., 2020). Other dyes, like malachite green and leuco-crystal violet, change their emitted colour upon interaction with LAMP products (Gachugia et al., 2020; Martín-Ramírez et al., 2022). Metal ion indicators like hydroxynaphthol blue (HNB), calcein, and Eriochrome black T produced visible colour changes perceived by the naked eye, through Mg²⁺ binding (Logeshwari et al., 2022; Park, 2022). Some indicators, including calcein and HNB, exhibit fluorescence when exposed to specific excitation wavelengths as well (Fischbach et al., 2015; Li et al., 2023).

Fluorescent dyes include DNA-intercalating compounds, such as SYBR Green I, Eva Green, Berberine, SYTO family (e.g., SYTO 9, 16, 13, 82, 81, and 84) (Fischbach et al., 2015; Quyen et al., 2019; Ruang-areerate et al., 2022). Some LAMP assays also use fluorophore-labelled primers that release fluorescence during amplification (Crego-Vicente et al., 2024).

LAMP is a promising diagnostic tool for pneumonia, offering high sensitivity and specificity, ease of use, low cost, and ease of result interpretation, making it suitable for point-of-care (POC) diagnostic (Soroka et al., 2021). Although LAMP has shown promise in detecting respiratory (Kang et al., 2012; Vergara et al., 2020; Lee et al., 2022), no existing LAMP-based screening panels currently target the most common bacterial causes of childhood pneumonia using a simple, easy-to-interpret assay. Most existing studies still rely on complex and expensive PCR-based equipment.

This study aimed to standardise an accessible, robust, and visually interpretable LAMP-based detection method for the direct identification of the main bacterial pathogens causing childhood pneumonia (S. pneumoniae, Staphylococcus aureus, H. influenzae, K. pneumoniae, and M. pneumoniae), without requiring advanced equipment or complex sample pre-treatment.

LAMP technology is widely used for pathogen detection due to its high sensitivity, specificity, simplicity, and rapid results (Atceken et al., 2023) meeting WHO criteria for tuberculosis diagnosis (WHO, 2016). Although it has been extensively applied to viral respiratory pathogens, especially after the SARS-CoV-2 pandemic, few studies focus on key bacterial agents of childhood pneumonia (Pneumonia in children: What you need to know | UNICEF, 2024). Many existing methods rely on complex detection systems, limiting use in high-incidence, low-resource settings.

This study describes a LAMP-based panel designed to detect key bacterial causes of childhood pneumonia, aiming to standardise a reliable, user-friendly method with straightforward interpretation. As a first step, we tested various dyes to enable naked-eye interpretation of LAMP results without specialised equipment.

SYBR Safe, an accessible dye that intercalates into dsDNA and fluoresces when excited at 280 nm or 502 nm, was initially evaluated (Kolbeck et al., 2021). This was demonstrated in the results, where positive reactions showed stronger fluorescence than negatives, but residual signals in negative tubes made interpretation difficult. This background may stem from SYBR Safe binding not only to dsDNA, but also to primers and single-stranded DNA (ssDNA), as reported for SYBR Green (Shaik et al., 2008; Dragan et al., 2012). Higher dye concentrations also reduced efficiency and product yield, leading us to discard them. Though SYBR Safe has been used in LAMP for Leishmania detection (Thita et al., 2019) visual interpretation remained challenging. Later, combining with gold nanoparticles improved clarity, but only in post-reaction steps (Ruang-areerate et al., 2021).

Since SYBR Safe lacked a clear visual distinction, we switched to calcein chelated with Mn²⁺. Calcein’s fluorescence is quenched by Mn²⁺, but during LAMP amplification, generated pyrophosphate binds and precipitates Mn²⁺, releasing fluorescence when excited near 480 nm (Mansour et al., 2015). Calcein also enables colour changes, slightly yellow for negative and green for positive (Tomita et al., 2008). Optimizing calcein-Mn²⁺ and Mg²⁺ concentrations is crucial to visually distinguishing LAMP results. In our study, adjustments had little impact on fluorescence or colour change. A persistent background signal in all tubes, including negatives, made interpretation difficult and risked false positives. Sensitivity was also low, with only high concentrations of S. pneumoniae showing a clear positive signal. This may stem from incomplete calcein-Mn²⁺ chelation, leaving residual fluorescence. Increasing Mn²⁺ reduces this background but also impairs reaction efficiency, likely due to Mn²⁺ competing with Mg²⁺, which the polymerase requires. While some polymerases can use Mn²⁺ as a cofactor, it often reduces performance and may destabilise important LAMP structures like loops or hairpins (Millonig et al., 2009; Vashishtha and Konigsberg, 2016).

Adjusting the [Mn²⁺: Mg²⁺] ratio can enhance contrast between positive and negative LAMP reactions. However, in our system, even at different ratios, the distinction remained unclear. Although a 4:1 ratio has been reported to improve visual differentiation (Petrone et al., 2015), residual fluorescence persisted in negative tubes, making result interpretation difficult.

We next evaluated SYTO 9, a dye that binds to dsDNA and emits strong green fluorescence when excited at 480–500 nm (Monis et al., 2005). Compared to SYBR Safe dyes, SYTO 9 offers less inhibition, stronger fluorescence, and a better signal-to-noise ratio, reducing false positives (Ma et al., 2019). However, when used alone, it failed to clearly differentiate positives from negatives by eye, as both emitted similar fluorescence. Although SYTO dyes have been used in LAMP, those studies typically relied on real-time systems with sensitive detectors to capture subtle differences (Chen and Ge, 2010; Patel et al., 2013; Oscorbin et al., 2016; Quyen et al., 2019). Fluorescence in negative tubes may result from SYTO 9 binding to primers or template DNA, even without amplification (Meagher et al., 2018). To address this, a dual-dye system was proposed, combining SYTO 9 to mark positives and a second dye to mask its background signal. SYTO 9 was paired with HNB as previously reported (Li et al., 2023). HNB emits reddish-orange fluorescence when interacting with Mg²⁺ under ~470 nm light. In positives, SYTO 9 binds to amplified dsDNA, emitting a strong green signal. In negatives, HNB masks any weak SYTO 9 fluorescence, offering a clear visual contrast: green for positives, reddish-orange for negatives.

We adopted this dye combination with modifications to optimise concentrations for our LAMP. HNB fluorescence was initially too faint to allow a clear visual distinction between positive and negative reactions. The optimal concentrations were 341.25 μM for HNB, 13 times higher than previously reported, and 0.75 μM for SYTO 9. Higher HNB intensified the red hue in negatives but gave positives a reddish-orange tone. Increasing SYTO 9 countered this, especially at low HNB levels, highlighting the need to fine-tune dye concentrations for each LAMP system.

Mg²⁺ concentration influences both enzyme activity and fluorescence intensity (Oscorbin and Filipenko, 2023), due to its interaction with HNB, which produces the reddish signal in negative tubes. We tested Mg²⁺ concentrations between 6 and 8 mM, assessing fluorescence and product yield. Higher Mg²⁺ enhanced amplifications but also strengthened HNB fluorescence, masking SYTO 9’s positive green signal and shifting colour to orange. Since Mg²⁺ affects HNB’s absorptions and emission spectra (Goto et al., 2009; Ding et al., 2015) we chose a 6 mM Mg²⁺. It provided clear green fluorescence in positives, a distinct red in negatives, and strong amplification.

Reaction temperature affects enzyme performance. At 65 °C, LAMP produced the most intense band pattern, aligning with its reported optimum (Oscorbin and Filipenko, 2023). Components like KCl, H₃PO₄, and Tris–HCl can dampen HNB’s reddish fluorescence, but this effect is less noticeable above 63 °C, even after cooling to 4 °C (Li et al., 2023). Running the reaction at 65 °C helps maintain a strong, clear HNB signal, supporting effective visual interpretation.

Both SYTO 9’s green fluorescence and HNB’s reddish signal remained stable for over 52 days, far longer than the 9 days previously reported (Li et al., 2023). Using higher dye concentrations and storing samples away from light likely preserved signal clarity, allowing greater flexibility for reading or checking results after amplification. Over time, SYTO 9’s green fluorescence faded to greenish-yellow. Although photobleaching was limited by dark storage, reversible binding to amplification products may also explain this event (Biebricher et al., 2015; Stiefel et al., 2015). In contrast, HNB fluorescence intensified, possibly due to stronger interactions with Mg²⁺ developing over time (Goto et al., 2009).

Under the established reaction conditions, the sensitivity assay showed a LoD of 3.9 ×10³ CFU/mL for S. pneumoniae, matching previous study using the same primers, which reported 10³ copies/mL (Kang et al., 2012). This aligns with other reports with LoD values of 10³ CFU/mL (de Paz et al., 2020; Si et al., 2021) using real time detection systems are usually more sensitive than the visual method. These results suggest the method’s suitability for use in low-resource settings.

The LoD for S. aureus was 1.7 ×10⁵ CFU/mL, aligning with a previous study using the same primers (10⁵ copies/mL) (Kang et al., 2012) and others that have reported between 10⁴ (Si et al., 2021) and 10⁵ CFU/mL by flanking the same gene (femA) (Vergara et al., 2020; Kim et al., 2022). For H. influenzae, our LoD of 8.2 ×10³ CFU/mL was slightly lower than the 10⁵ copies/mL reported (Kang et al., 2012). This difference likely reflects methodological variations, as our assay used bacterial suspension at known concentrations rather than clinical cut-offs. Overall, our findings are consistent with reported LoDs of 10³ CFU/mL (Kim et al., 2011; Takano et al., 2017; Wang et al., 2022).

As shown in Table 4, a clinical sample positive for S. aureus with a bacterial load of 4 ×10⁴ CFU/mL was successfully detected by LAMP, despite being below the reported LoD. This may reflect differences between controlled suspensions and clinical matrices, which can affect DNA availability; for instance, some bacteria in clinical samples may be lysed, releasing DNA and facilitating detection (Ren et al., 2020). Additionally, the LoD is an estimated threshold, and stochastic amplification can occasionally yield positive results below this value.

In M. pneumoniae, the LoD detected in our study (1.27 ×10³ genome copies/reaction) was higher than previously reported (Petrone et al., 2015) likely due to differences in the readout of the reaction. While droplet digital PCR (ddPCR) can detect as few as 2.9 copies/reaction of M. pneumoniae (Zhao et al., 2023) LAMP is simpler and easier to perform. Given that clinical loads above 10⁵ copies/mL indicate active infection, the sensitivity achieved with the standardised LAMP assay remains appropriate for detecting this bacterium (Liu et al., 2019; Shi et al., 2019; Zhao et al., 2020).

An initial primer set for K. pneumoniae described previously failed to produce adequate amplification under our reaction conditions (data not shown). Therefore, a new set was designed targeting the khe gene encoding haemolysin. Alignment of 502 nt sequences of this gene showed high conservation, with only three variations, increasing the likelihood of detecting clinically relevant strains. In silico analysis showed high specificity to K. pneumoniae complex, and through in vitro analysis, no cross-reactivity was detected with other bacteria, including closely related species, confirming the analytical specificity and reducing the risk of false positives (Yin-Ching et al., 2002; Feng et al., 2023). This confirms the specificity and reliability of the primer set for accurate detection of K. pneumoniae.

The LoD obtained with the designed primers was 1.5 ×10⁴ CFU/mL, which is consistent with other reports with LoDs of 10⁴ (Vergara et al., 2020) and 10⁵ CFU/mL (Qiu et al., 2022). However, some studies have reported LoDs near or below 10 CFU/mL (Banerjee et al., 2024; Bermúdez-Fornos et al., 2025).

Accurate diagnosis should not rely solely on rapid tests like the one proposed, but must consider the patient’s clinical context. Interpreting results alongside estimated bacterial load can help distinguish true infection from colonisation as higher concentrations are generally seen during active infections due to failed immune regulatory mechanisms (Beisswenger et al., 2009; Leshem et al., 2020; Woelfel et al., 2024).

Since molecular tests amplify faster at higher target concentrations (De Arcos-Jiménez et al., 2025), we proposed that the time to positivity in our reaction could provide an additional indicator to differentiate infection from colonisation in positive cases.

In our study, the absence of a positive signal at 45 min for S. pneumoniae, S. aureus, and at 50 min for H. influenzae may suggest bacterial concentrations below 10⁵ CFU/mL. This threshold is relevant for several reasons.

Various studies reported high S. pneumoniae loads in confirmed pneumonia cases (Dubot-Pérès et al., 2025) often exceeding 10⁵ CFU/mL in the nasopharynx (Vu et al., 2011) with 10⁶ CFU/mL suggested as a threshold to distinguish infection from colonisation (Brotons et al., 2017; Haddar et al., 2020). Similarly, S. aureus loads tend to be higher during active infection, with 10⁵ CFU/mL proposed as a diagnostic cut-off for methicillin-resistant S. aureus (MRSA) (Kwon et al., 2012). Other studies found pneumonia cases averaging 10⁷ copies/mL, while colonised patients remain below 10⁵ copies/mL, supporting its use to differentiate infection from colonisation (Huang et al., 2015).

Haemophilus influenzae loads above 10⁴ CFU/mL have been linked to active infection (Hare et al., 2018). Concentrations exceeding 10⁶ copies/mL are associated with confirmed childhood pneumonia (Park et al., 2017) and thresholds of >10⁵ CFU/mL may help indicate an ongoing infectious process (Kais et al., 2006).

Considering that, bacterial loads above 10⁵ CFU/mL are associated with active infections, for S. pneumoniae and S. aureus, a cut-off time of 45 min could help differentiate colonisation from true infection in our LAMP reaction. The cut-off point for H. influenzae should be set at 50 min, based on the association of loads above 10⁴ CFU/mL with active infections. Therefore, the time to positivity in our reaction may serve as a useful marker to distinguish between colonisation and infection. The use of bacterial suspension provides a controlled framework to assess the assay; however, validation with clinical samples remains necessary to confirm its relevance in real clinical settings.

Evaluation of the LAMP assay with clinical samples showed that the reaction can be successfully performed even with templates exhibiting variability in composition and physical characteristics. While the study primarily aimed at standardising the detection technique rather than a full clinical validation, it was important to confirm applicability to real samples, hence the limited number of specimens tested. Interestingly, in a couple of cases, LAMP detected an additional bacterium beyond that identified by culture or PCR, suggesting higher sensitivity and the ability to detect bacteria present at low load or affected by prior empirical treatment (Bouzada et al., 2025).

As previously mentioned, childhood pneumonia can be caused not only by bacteria but also by viruses, including RSV, HRV, hMPV, and influenza viruses, and others. In some cases, fungal pathogens like P. jirovecii may be responsible (Pneumonia in children, 2022). However, the present study focuses exclusively on bacterial detection, as these pathogens can cause more severe infections in the paediatric population (Nascimento-Carvalho et al., 2016; Nathan et al., 2020). Future adaptations and improvements of the proposed technique could include non-bacterial pathogens or even combined panels, given the potential of coinfections.

The LAMP technology described here meets most REASSURED criteria (real-time connectivity, ease of specimen collection, affordable, sensitive, specific, user-friendly, rapid and robust, equipment-free or simple, deliverable) (Land et al., 2019). However, an important limitation of this study is the need for further validation using clinical samples before confident diagnostic application.