The use of light as an alternative to chemical or thermal inhibition of unwanted microorganisms has received increased attention in medical, food and agricultural, water, and hygiene applications. The germicidal impact of UV-light is well established (Hijnen et al., 2006; Koutchma, 2008, 2018; Hong et al., 2014; Song et al., 2016; Marasini et al., 2021), especially with respect to microbes associated to surfaces. It has been successfully used in controlling water disinfection (e.g., Pullerits et al., 2020), various skin infections (e.g., Abana et al., 2017) as well as surgical site infections (Buonanno et al., 2013). Modification of the wavelength within the UV-band from 256 to 207 nm reduces the negative impact on human cells while maintaining the germicidal effect as stated for methicillin resistant Staphylococcus aureus (Buonanno et al., 2013). Despite a high interest in UV-treatment in food contexts to prolong shelf-life and increase food safety when thermal treatment is not an option, legislation varies in different countries. As such, implementation of UV-light and novel food ingredients requires approval in response to various regulations (Koutchma, 2018). Implementation of UV-based disinfection treatment must take occupational environment legislation into account. Given these disadvantages and corroborated by the development and low cost of light-emitting diodes (LED), applications in the visible light spectrum (400–760 nm) have been highlighted (Lui et al., 2016). This is especially true for the blue, green and red spectral bands.

Encouraging results have repeatedly been reported upon photobiological treatment of hazardous microorganisms using blue light (400–500 nm) for different applications, e.g., food safety control (Neal et al., 2012; Guffey et al., 2016; Lacombe et al., 2016; Josewin et al., 2018; Ghate et al., 2019; Hyun and Lee, 2020; Hyun et al., 2021), microbial disease treatment (Plavskii et al., 2018), esp., multidrug resistant strains (Ferrer-Espada et al., 2018), odontology (Cieplik et al., 2014), dermatology (Maclean et al., 2009) or sewage treatment (Giannakis et al., 2015). Some non-phototrophic bacteria are equipped with blue light receptor protein domains (Alsanius et al., 2019 and references therein). The physiological response to blue light varies between different microorganisms and their nutritional environment (Gharaie et al., 2017; Alsanius et al., 2021). Basic functions for microbial spreading and establishment on surfaces may be affected by blue light exposure (Vermeulen et al., 2008; Maclean et al., 2009; Alsanius et al., 2021). Suggested mechanisms involved into the photoinhibitory effect of blue light revolve around the presence of photoreceptor proteins, such as porphyrins or flavins, that can capture a photon converting it into an electronic signaling mechanism (Krauss, 2007). For disinfecting purposes, this signaling mechanism can be used against bacteria through a cascading chain of events ultimately leading to bacterial inhibition or mortality. In the presence of the proper wavelength of light at high enough intensities, Reactive Oxygen Species (ROS) are generated due to this photon capture and conversion, ultimately causing damage to the cell wall (Ryter et al., 2007), DNA, proteins and lipids. Possible photoinhibitory effects are summarized in Figure 1.

In this systematic review, we consider the impact of light quality (i.e., the spectral distribution of light) in the short-waved blue band on Escherichia coli (E. coli). We focus on the following research questions:

The aggregated results from the initial random effects meta-analysis of all included studies indicated a significant decrease of CFU after blue light treatment. The aggregated effect size (g) was 1.49 (95% confidence interval: 0.64–2.34; Z = 3.43, p = 0.0006; χ² = 147.19, p < 0.00001; I² = 95%). Estimated effects for each included study are presented in Figure 3 and Table 2. As expected, similar aggregated results were found in the subsequent subgroup-analyses.

The subgroup-analysis of intensities indicated a significant decrease of CFU after blue light treatment for all studies using intensities of 5.52–50 mW/cm², but none of the studies using intensities of 60–1,262.6 mW/cm². Differences in CFU decrease among subgroups were significant (χ² = 135.5, p < 0.00001). Tests of statistical heterogeneity for subgroup 60 mW/cm² showed χ² = 0.23, p = 0.64; I² = 0%. Tests of statistical heterogeneity were not applicable to the remaining subgroups of this analysis as the amount of studies was insufficient. Statistical heterogeneity for the entire intensities subgroup-analysis was indicated by χ² = 135.71, p < 0.00001; I² = 93%. Figure 4 and Table 3 display estimated effects for each included study and subgroup.

In the subgroup-analysis of wavelengths, significant CFU decrease was seen in subgroups of 395–445 nm but not in subgroups of 455–460 nm. Differences in CFU decrease among subgroups were significant (χ² = 83.6, p < 0.00001) Tests of statistical heterogeneity for subgroup 405 nm showed χ² = 37.23, p < 0.00001; I² = 84%, and χ² = 28.22, < 0.00001; I² = 96% for subgroup 460 nm. Tests of statistical heterogeneity were not applicable to the remaining subgroups of this analysis as the amount of studies was insufficient. Tests of statistical heterogeneity for the entire wavelengths subgroup-analysis showed χ² = 155.03, p < 0.00001; I² = 92%. Figure 5 and Table 4 display estimated effects for each included study and subgroup.

The subgroup-analysis of exposure dose indicated scattered results, where a significant CFU decrease was seen among subgroups of exposure dose 10, 30, 45, 60, 70, 90, 100, 135, 180, 240, 360, 480 and 540 J/cm² but not among subgroups of exposure dose 0.3312, 0.6624, 1.3248, 1.9872, 2.6496, 3.312, 25, 50, 120, 150, 162, 216 or 576 J/cm². Differences in CFU decrease among subgroups were significant (χ² = 105.5, p < 0.00001). Tests of statistical heterogeneity showed following results: χ² = 0.03, p = 0.85; I² = 0% for subgroup 10 J/cm²; χ² = 0.00, p = 0.99; I² = 0% for subgroup 30 J/cm²; χ² = 7.87, p = 0.005; I² = 87% for subgroup 50 J/cm²; χ² = 5.09, p = 0.02; I² = 80% for subgroup 180 J/cm²; χ² = 2.11, p = 0.15; I² = 53% for subgroup 360 J/cm². Tests of statistical heterogeneity were not applicable to the remaining subgroups of this analysis as the amount of studies was insufficient. Tests of statistical heterogeneity for the entire exposure dose subgroup-analysis showed χ² = 165.25, p < 0.00001; I² = 81%. Figure 6 and Table 5 display estimated effects for each included study and subgroup.

When analyzing initial and final CFU based on serovar/pathovar in the last subgroup-analysis, significant CFU decrease was seen for ATCC8793, O157H7, DH5a and K12 but not for AF0006, ATCC 25922, E343, E402, E9034A, EC958, MG1655 or UT189. Differences in CFU decrease among subgroups were significant (χ² = 61.1, p < 0.00001). Tests of statistical heterogeneity showed following results: χ² = 5.65, p = 0.02; I² = 82% for subgroup O157H7; χ² = 0.06, p = 0.81; I² = 0% for subgroup UT189; χ² = 18.37, p = 0.0001; I² = 95% for subgroup ATCC 25922; χ² = 9.02, p = 0.003; I² = 89% for subgroup K12. Tests of statistical heterogeneity were not applicable to the remaining subgroups of this analysis as the amount of studies was insufficient. Tests of statistical heterogeneity for the entire serovar/pathovar subgroup-analysis showed χ² = 114.39, p < 0.00001; I² = 87%. Figure 7 and Table 6 display estimated effects for each included study and subgroup.

Apart from the broad diversity of light intensities and exposure as well as E. coli strains, a range of matrices were used in the present set of studies, complicating comparability between studies. The overall outcome quality assessment of bias indicated in general some concerns or high risk of bias across studies (Figure 8). In detail, none of the studies was distinct with respect to randomization of samples during preparation, exposure, nor during analysis (order of sampling, enumeration; domain 1 “Randomization,” D1). An additional confounding factor was the absence of standardization procedures with respect to the physiological status of the inoculum (exceptions: Abana et al., 2017; Kim and Kang, 2021) or ray distribution (exception: Hoenes et al., 2015). No deviations from intended interventions or withdrawal of information were observed (Domain 2, D2). In this context it is worthwhile to note, that the studies only report on sample numbers included into the statistical analysis and thus do not allow a clear judgment. Likewise, with respect to domain 3, missing outcome data (D3), the assessment was left with uncertainties. Multiple concerns became evident with respect to the measurement of the outcome (domain 4, D4). Most studies based their analysis on viable count. Upon stress, cells of some microbial species may transcend to different forms of dormancy, e.g., viable but not culturable (VBNC), persister cells. This has also been demonstrated for E. coli. Only two of the involved studies (Abana et al., 2017; Kim and Kang, 2021) account for the dilemma evolving from the exclusiveness in viability concepts, but these studies used different methods to assess viability. There are several additional issues regarding measurement—namely absence of true control treatments (unexposed samples) as well as uncertainties regarding abiotic experimental conditions (i.e., temperature) and exposure length. In combination, they may provoke a crucial mass of biases. (i) the interaction between microbial growth and temperature is well known. Temperature regime during inoculum propagation and during blue light incubation influences cell viability. (ii) Exposure is a function of light intensity and time (Equation 3). Exposure length may be used to compensate for light sources with lower intensity. Depending on the strength of the light source, the additional exposure length may be substantial and additional microbial propagation may take place. It is thus critical that true controls are incorporated into such experiments, allowing standardization of potential additional growth. The present set of studies used the microbial titer before exposure to blue light as the reference to compare the outcome of the treatment. This complex of biases needs to be viewed in the light of lacking reports on experimental conditions and cell age. Last, but not least, concerns with respect to the selection of reported results (domain 5, D5) need attention, as few studies distinctly document or involve enough true replicates, technical replicates and repetitions of the interventions. Ultimately, no studies were selected for removal due to bias because any potential discriminating factors were widespread throughout studies on blue light decontamination of E. coli.

There is a substantial interest from medical, food and environmental sciences on short waved blue light to circumvent the use of antibiotics to treat bacterial skin and oral infections, to increase food safety, and to aid in environmental remediation (e.g., sewage water). Blue light often appears to be the new silver bullet. Indeed, this systematic review supports the reducing potential by the short wave blue light of culturable E. coli (Figure 3). From the included body of literature, both light intensity and wavelength were identified as the most decisive factors to reduce the number of culturable E. coli. Despite the significantly large effect size, considerable heterogeneity was present across studies, meaning that the effect of blue light treatment is not uniform. The impact of light intensities and wavelengths on the occurrence of culturable E. coli ceased when exceeding 50 mW/cm² and 445 nm, respectively. However, caution is required when interpreting these results. For example, only the intensity subgroup employing 60 mW/cm² consisted of more than one study. A light intensity of 50 mW/cm² might be considered as a possible cut-off value. However, the considerable heterogeneity across all studies blurs the picture and significant results were obtained for some but not all subgroups. The same applies to the subgroup analysis with respect to wavelengths. The existing body of information does not allow any recommendations on suitable intensity, dose or exposure of blue light toward E. coli for sanitation or cure, as the number of eligible studies is low and many studies.

There is a considerable weakness to the mode in which the inactivation of E. coli was assessed in the included studies. In all studies, viable count through enumeration of colony-forming units was used. This is a well-established method to estimate the presence of bacteria and bacterial groups which can grow on stationary microbiological media, but there are inherent pitfalls to this method. For the interpretation of the given information, the detection limit needs to be defined. “No growth” on stationary medium under standardized conditions is often interpreted as “absence.” Apart from their potential presence below the detection limit, cells might be present as viable but non-culturable cells (VBNC) and persister cells. The transition to the viable but not culturable stage has been shown for several organisms, including E. coli (Oliver, 2010). Upon hostile environmental conditions, cells thus switch to a low, but measurable metabolic stage with maintained respiration and cell wall integrity, but can retain virulence and infectivity upon resuscitation (Oliver, 2010). Bacterial subpopulations entering a dormant, non-dividing stage upon environmental stress (e.g., chemotherapy) are called persister cells. They are known to survive conditions lethal to regular cells (Lewis, 2010). Therefore, revitalization might occur as shown in Giannakis et al. (2015). All three post exposure options listed here lead to recolonization or recurrent infection. The included body of studies most likely underestimate the number of living cells and thus the level of inactivation. RNA-based approaches involving RT-PCR or ddPCR might give a more distinct result on the blue light impact on E. coli.

Information on the cell culture’s physiological status is vital and might account for some of the conflicting or scattered results. As demonstrated by Abana et al. (2017) sensitivity to blue light varies between different physiological stages (exponential, transition, stationary). For four out of eight tested strains, the strongest sensitivity to blue light exposure occurred during the exponential phase, while 25% of the tested strains displayed strongest sensitivity during stationary phase when compared to the unexposed controls. Likewise, the mode of growth (planktonic, biofilm) needs to be taken into account, as marginal blue light impact was observed for biofilm associated E. coli (Ferrer-Espada et al., 2018, 2020). Thus display of cardinal growth parameters (e.g., generation time) and tested physiological stage for tested strains, their mode of growth as well as a robust operation parameters during propagation and exposure must be included to find relevant PDI levels.

CL, SW, and BA: conceptualization, methodology, validation, formal analysis, investigation, writing—original draft preparation, writing—review and editing, and visualization. BA: resources, project administration, and funding acquisition. CL: data curation. All authors contributed to the article and approved the submitted version.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.