The in situ bacterial production of fluorescent organic matter; an investigation at a species level.

Aquatic dissolved organic matter (DOM) plays an essential role in biogeochemical cycling and transport of organic matter throughout the hydrological continuum. To characterise microbially-derived organic matter (OM) from common environmental microorganisms ( Escherichia coli, Bacillus subtilis and Pseu- domonas aeruginosa) , excitation-emission matrix (EEM) ﬂ uorescence spectroscopy was employed. This work shows that bacterial organisms can produce ﬂ uorescent organic matter (FOM) in situ and, furthermore, that the production of FOM differs at a bacterial species level. This production can be attributed to structural biological compounds, speci ﬁ c functional proteins (e.g. pyoverdine production by P. aeruginosa ), and/or metabolic by-products. Bacterial growth curve data demonstrates that the production of FOM is fundamentally related to microbial metabolism. For example, the majority of Peak T ﬂ uorescence ( > 75%) is shown to be intracellular in origin, as a result of the building of proteins for growth and metabolism. This underpins the use of Peak T as a measure of microbial activity, as opposed to bacterial enumeration as has been previously suggested. This study shows that different bacterial species produce a range of FOM that has historically been attributed to high molecular weight allochthonous material or the degradation of terrestrial FOM. We provide de ﬁ nitive evidence that, in fact, it can be produced by microbes within a model system (autochthonous), providing new insights into the possible origin of allochthonous and autochthonous organic material present in aquatic systems.


Introduction
Dissolved organic matter (DOM) in aquatic systems plays an essential role in global biogeochemical cycling (Bieroza and Heathwaite, 2016;Hudson et al., 2007). It is generally accepted that the majority of DOM found in freshwaters is allochthonous, with a proportion of the DOM considered to be produced in situ, i.e. autochthonous material (Coble et al., 2014). Fluorescence excitation-emission matrix (EEM) spectroscopy has been increasingly employed in recent research to characterise aquatic fluorescent organic matter (FOM) and fluorescent dissolved organic matter (FDOM) (Baker, 2005;Bridgeman et al., 2015). The use of this technique has advanced our understanding of FDOM, its classification, transformation and potential origin (Hudson et al., 2007;Stedmon and Bro, 2008).
Aquatic FDOM has been characterised as consisting of humiclike material considered to be of allochthonous origin of terrestrial input (Coble et al., 2014). The compounds associated with terrestrially derived FDOM are known to be stable higher molecular weight aromatic compounds, generally considered non-labile (Cooper et al., 2016). However, recent work concerning the marine environment has suggested that humic-like FDOM could be a consequence of bacterial metabolism (Guillemette and del Giorgio, 2012;Kramer and Herndl, 2004;Romera-Castillo et al., 2011;Shimotori et al., 2012). Recent findings by Kallenbach et al. (2016) have shown the production of extracellular humic material by bacteria within soil organic matter. There is no direct evidence that the production of humic-like FDOM in freshwaters is the result of bacterial processing. However, Elliott et al. (2006) attributed the presence of this FOM in laboratory samples to stress as opposed to a function that may inherently occur within aquatic systems. What is clear from the literature is that a more detailed understanding of microbial/OM interactions in freshwater systems is needed.
This protein-like FDOM is attributed and assumed to be of microbial origin (Cammack et al., 2004;Coble et al., 2014;Hambly et al., 2015;Smith et al., 2004). Recent literature suggests that Peak T fluorescence may act as a surrogate for microbial and bacterial activity Cumberland et al., 2012), as first highlighted by Hudson et al. (2008). Recent surface freshwater research has also attempted to use Peak T fluorescence to determine enumeration of specific species. For example, Baker et al. (2015) observed a log correlation R ¼ 0.74 across a 7-log range in Escherichia coli enumeration for sewage impacted rivers. Using Peak T fluorescence to infer microbial enumeration, and activity, has been further suggested for groundwater systems, where there is little background fluorescence interference (Sorensen et al., 2016(Sorensen et al., , 2015. Sorensen et al. (2015) investigated low levels of microbial contamination in drinking water supplies, reporting linear correlations, R 2 ¼ 0.57 from < 2 to 700 cfu 100 ml À1 . Although relationships have been demonstrated for protein-like fluorescence and the presence of bacteria in freshwater systems, the research reported thus far does not take into account the implication and impact of microbial activity at an individual species level.
The study aim was to further our understanding of the role aquatic microbes play in the production of both protein-like and humic-like FOM in freshwaters. For this, we focus on the development of FOM in a model system using a simplified microbial community, thus removing the background complexities observed in environmental samples. Using this approach, we also determine the intracellular and extracellular fluorescence signatures of common freshwater bacterial species.

Bacterial species
Three bacterial species were cultured for analysis; Escherichia coli (ATCC 10536) was used as its presence in freshwaters can indicate sewage contamination (Sigee, 2004); Bacillus subtilis (ATCC 6633) was used as it is a ubiquitous soil bacterium (Graumann, 2007) that may be transferred into freshwater systems; and Pseudomonas aeruginosa (NCIMB 8295) as it is ubiquitous in freshwater systems (Elliott et al., 2006;Sigee, 2004).

Media
A non-fluorescent minimal media was developed to promote growth within our model system whilst excluding the presence of proteinaceous material. The basal medium consisted of a final concentration of 0.2% v/v glucose solution, as the sole carbon source, and a solution containing a source of phosphate, nitrogen, sodium and magnesium. The basal medium was adopted from the ATCC ® medium 778 Davis and Mingioli minimal medium (Davis and Mingioli, 1950), but without the addition of amino acids and agar. All elements of the basal medium were filter sterilised using a Minisart ® 0.2 mm cellulose filter (Sartorius Stedim Biotech, Germany). CaCl 2 (final concentration 0.035% v/v) and a trace element solution (final concentration 0.1% v/v), obtained from Kragelund and Nybroe (1994), were added to the sterile basal medium prior to inoculation. These chemicals were sterilised by autoclaving at 121 C for 15 min.

Fluorescence measurements
Fluorescence excitation-emission matrices (EEMs) were collected using an Aqualog ® (Horiba Ltd., Japan). Samples were not filtered prior to fluorescence spectroscopic analysis (except for bacterial supernatant samples, section 2.6). The scan parameters employed were; excitation wavelengths from 200 to 500 nm via 1 nm steps, and emission wavelengths of 247.88e829.85 nm in 1.16 nm steps using an integration time of 500 ms. A micro quartz cuvette (1400 mL) with a 10 mm path-length was used throughout.
Spectra were blank subtracted, corrected for inner filter effects (for both excitation and emission wavelengths) and first and second order Rayleigh Scattering masked (±10 nm at l ex ¼ l em and 2l ex ¼ l em ) (Coble et al., 2014;McKnight et al., 2001). Fluorescence data is reported in quinine sulphate units (QSU), determined from normalising data to the fluorescence from 1 mg L À1 quinine sulphate at l ex ¼ 347.5 nm and l em ¼ 450 nm (Kramer and Herndl, 2004;Mostofa et al., 2013;Shimotori et al., 2012Shimotori et al., , 2009. Instrument validation was undertaken daily with a quinine sulphate standard (Starna Cells, USA), with CV being < 3% (n ¼ 5) in all events.

Fluorescence data analysis
A custom script, written in Python™ (Python Software Foundation), was used to convert the data into QSU and create the EEM maps. The script crops the data window to l ex 240e490 nm, l em 250e550 nm to allow for the analysis of the UV spectra, the area of interest within FDOM work. Data l ex < 240 nm was discounted due to the data quality produced by the Aqualog ® caused by the signal to noise ratio. The custom script was then used to undertake peak picking for specific fluorescence peaks. Some of the peak picked data was normalised to the maxima to provide a clear visual representation of the fluorescence development over time. The EEM data was also investigated by employing parallel factor (PARAFAC) analysis (Stedmon and Bro, 2008) in Solo (Eigenvector Research Inc., WA, USA) software, in conjunction with the MATLAB ® PLS-Toolbox (Mathworks, USA).

Bacterial growth curves
Growth curves (n ¼ 9 i.e. nine independent replicates) of each bacterial species were undertaken by inoculating 150 mL of the sterile medium (section 2.2) from a fresh overnight plate culture (< 24 h) and incubating the samples at 37 C, shaking at 150 rpm. Aliquots were collected every 30 min for fluorescence measurements (section 2.3) and optical density (OD) measurements at 600 nm (WPA Spectrawave S1200, Biochrom, UK); OD, attenuation determined by absorbance and scattering, is routinely used to represent the relative increase in cell numbers within a sample when monitoring bacterial growth (Hall et al., 2014). OD data was also normalised to the maxima.

Bacterial culture analysis
Media was inoculated, from a fresh overnight plate culture (< 24 h), with each of the bacterial species and incubated overnight at 37 C, shaking at 150 rpm throughout. Overnight cultures were centrifuged at 5000Âg for 5 min (Allegra X-30R, Beckman Coulter™, USA) to form a bacterial pellet. The supernatant was pipetted off and filtered using a Minisart ® 0.2 mm cellulose filter (Sartorius Stedim Biotech, Germany) to guarantee all cells were removed. The pellet was resuspended and washed 3 times in 5 mL of ¼ strength Ringer solution (Oxoid Ltd., UK) to ensure that any supernatant or media was no longer present. To physically lyse the cells, a 1 mL aliquot of the resuspended cells was sonicated (Ultrasonic Processor XL 2020, Misonix Inc., US) in three 10 s pulses at a fixed frequency of 20 KHz, not exceeding 40% amplitude, and kept over ice throughout (Doron, 2009). Physical lysis was undertaken to ensure no extra chemicals were added to the cells that may alter the fluorescence properties of the sample (nine independent replicates). An endospore suspension for B. subtilis was prepared as described by Lawrence and Palombo (2009). To check for the presence of endospores and the removal of vegetative cells, an endospore stain was conducted using the Schaeffer-Fulton method (Schaeffer and Fulton, 1933).

Results and discussion
Filtering of samples was not performed prior to spectroscopic analysis to maintain sample integrity , since the focus of this study is on in situ bacterial production of FOM in a model system. Each individual bacterial species exhibited unique fluorescing signatures. Some FOM, specifically Peak T, was dominant in all samples exhibiting high fluorescence intensities. This limited the application of PARAFAC analysis, whereby no robust model, CORCONDIA > 90% (Bro and Kiers, 2003), that adequately Fig. 1. Fluorescence and optical density (OD 600nm ) data for Escherichia coli growth curve, showing: a) optical density and fluorescence (QSU, 1 QSU ¼ 1 mg À1 quinine sulphate) ± 1 standard deviation (n ¼ 9); b) optical density and fluorescence data normalised to the maximum value ±1 standard deviation (n ¼ 9); c) excitation-emission matrix at time zero; and d) excitation-emission matrix at 360 min. explained the dataset could be identified. Subsequently, peak picking (Asmala et al., 2016), an established method for spectral analysis, was applied to peaks identified within the EEMs.

Bacterial growth curves
The E. coli growth curve is shown in Fig. 1, whereby Peak T is the dominant fluorescence peak and is present at time zero, upon initial addition of the E. coli cells (Dartnell et al., 2013;Sohn et al., 2009). During the growth curve, the intensity of Peak T increases in line with the optical density (OD) of the sample (Fig. 1). During the exponential stage (growth phase after acclimatisation; Hogg, 2005) there is a log increase in the intensity of Peak T fluorescence. This, alongside the increase in OD, leads to a significant strong correlation between Peak T and OD, R 2 ¼ 0.9821 (p < 0.001). This suggests that Peak T fluorescence intensity can be attributed to an increase in E. coli population size, in accordance with previous studies Cumberland et al., 2012;Dartnell et al., 2013;Deepa and Ganesh, 2017;Sohn et al., 2009). However, as tryptophan is an essential amino acid, necessary for protein formation during growth and other metabolic pathways, it will be produced as a result of cell multiplication and metabolic processing (Coble et al., 2014;Hogg, 2005). As such, Peak T fluorescence can also be attributed to E. coli cell activity. Fig. 1 shows that Peak C also develops during the exponential phase of the growth curve, exhibiting a lag in relation to the OD. The intensity of Peak C continues to increase even during stationary phase, in which cell deaths are equal to newly formed cells (Elliott et al., 2006;Hogg, 2005). Nevertheless a positive correlation between OD and Peak C fluorescence intensity is identified, R 2 ¼ 0.8624 (p < 0.001), supporting the association of Peak C with bacterial numbers. However, the observed lag in conjunction with the continued increase in fluorescence intensity during the stationary phase strongly supports the idea that metabolic activity, and not bacterial numbers per se, may be the main driver for the creation and production of Peak C fluorophores. Notably, the observed maximum fluorescence intensity of Peak C is a factor of 10 lower than Peak T (Fig. 1a). It can, therefore, be suggested that Peak C may be derived as a metabolic by-product or a secondary metabolite produced mainly during the stationary phase (Fig. 1b). Peak X (Table 1) is only present within the stationary phase, albeit at comparatively low fluorescence intensities (~30 QSU). The microbial production of Peaks C and X demonstrates the ability of E. coli to rapidly produce (within 8 h), in situ, FOM associated with allochthonous high molecular weight FOM. Fig. 2 highlights Peak T as the dominant fluorescence peak within the B. subtilis growth curve. Peak T intensity increases by an order of magnitude throughout the growth curve, in line with the increased OD (Fig. 2), demonstrating a strong significant correlation, R 2 ¼ 0.9879 (p < 0.005). However, as Peak T fluorescence intensity increases during what appears to be early stationary phase (Fig. 2), it could be suggested that these fluorophores are produced by metabolically active cells. This emphasises the use of Peak T as an indicator of microbial activity rather than being attributed to cell enumeration, despite the significant correlation identified. The production of Peak T within the stationary phase (Fig. 2) could also be related to B. subtilis sporulation, demonstrated by the high intensity Peak T fluorescence obtained bacterial endospores analysed alone (Fig. 3). This suggests that some of the fluorophores attributable to Peak T fluorescence are related to structural proteins since endospores are not metabolically active, although this is species specific.

Bacillus subtilis
Within the B. subtilis growth curve, Peak C demonstrates a sudden rise, at 360 min, prior to OD and Peak T development (Fig. 2), with a strong positive correlation between Peak C fluorescence intensity and the OD being identified, R 2 ¼ 0.9465 (p < 0.005). This further challenges our current understanding of Peak C being attributed to terrestrial allochthonous material (Coble et al., 2014). Smith et al. (2004) suggested that the Peak C fluorescence, identified in the presence of Bacillus sp., may be related to the fluorescence of endospores. However, Fig. 3 demonstrates the endospore suspension obtained from B. subtilis within our study as having high Peak T, and low Peak C, fluorescence intensity.
Florescence Peaks M and A M are produced and observed at very low intensities within the early stationary phase of the growth curve. A possible explanation of this observation is the result of the biodegradation of OM responsible for Peak C fluorescence, a process that has been noted in the literature (Coble et al., 2014). Alternatively, this could indicate that Peaks M and A M can be produced directly, in situ, by bacteria, as has been suggested to occur within marine environments (Coble, 1996;Shimotori et al., 2009).

Pseudomonas aeruginosa
Peak T is ubiquitous within the P. aeruginosa growth curve (Fig. 4), increasing by an order of magnitude within the exponential phase. A relatively weaker correlation, R 2 ¼ 0.7601 (p < 0.005), is identified between Peak T and the OD, likely to be caused by the upregulation of Peak T independent of cell number which can be seen at 330 min, in the late exponential, early stationary phase (Fig. 4). Prior to this, the Peak T fluorescence development tracks the OD (R 2 ¼ 0.9674, p < 0.05). One possible explanation for this sudden increase in Peak T fluorescence intensity is the production of exotoxin A; exotoxin A is an iron-scavenging enzyme that is produced by P. aeruginosa upon entry into stationary phase (Lory, 1986;Somerville et al., 1999). Previous studies have shown how Exotoxin A can be used to determine protein activity, by assessing tryptophan (Peak T) fluorescence quenching upon binding of NADþ to the enzyme active site Merrill, 1999, 1996;. Therefore, the observed subsequent sudden  decline in Peak T fluorescence intensity at 450 min may be as a result of this quenching phenomena (Fig. 4). P. aeruginosa has the most complex EEM spectra of the species analysed within this study (Dartnell et al., 2013;Elliott et al., 2006;Smith et al., 2004), with Peaks T, C and A C all immediately identified upon inoculation and during the lag phase (a period of acclimatisation; Hogg, 2005). The occurrence of Peaks C and A C at inoculation suggests it is likely that this FDOM is intracellular and produced within the cells during the initial overnight incubation; likely to be structural or functional proteins produced via microbial metabolic pathways, or potentially intracellular metabolic byproducts. These peaks increase log-fold throughout the growth curve, with both Peaks C and A C being correlated with, despite a lag in relation to, the OD; R 2 ¼ 0.7024 (p < 0.005) and R 2 ¼ 0.7146 (p < 0.005) respectively. The data indicates upregulation of these peaks during late exponential phase and stationary phase, suggesting that these peaks are a result of metabolic activity.
Peak Cþ develops rapidly and to a high intensity during the stationary phase of P. aeruginosa growth (Fig. 4) and this fluorescent peak is associated with the siderophore pyoverdine (Dartnell et al., 2013;Wasserman, 1965). Pyoverdine is an extracellular ironscavenging metabolite produced by P. aeruginosa and is associated with microbial virulence (da Silva and de Almeida, 2006). The fluorescence intensity of this high molecular weight OM within the P. aeruginosa growth curve, suggests that this Peak Cþ fluorescence could be derived from the building and exporting of pyoverdine. Peak Cþ has been seen in freshwater environments and is currently attributed to terrestrial allochthonous OM. However, our work proves that microbial compounds produced in situ (akin to autochthonous material) may contribute to this Peak Cþ fluorescence. As such, Peak Cþ may act as a biomarker for an active P. aeruginosa community, although further investigation within natural environmental systems is required.

Overnight culturing of bacterial species
From the microbial growth curve data it has been shown that all the fluorescence peaks identified (Table 1) are microbially produced in situ, with variations in peak occurrence between bacterial species. To further investigate the microbial source and origin of the OM, overnight cultures of each species were analysed to determine the presence of FDOM in the supernatant, OM within resuspended cells and lysed cells (see section 2.6). This provides a preliminary understanding of where the observed fluorescence is located post FOM production.
Peak T fluorescence is the only ubiquitous fluorescence peak common to all bacterial species cultured overnight (Table 2). This shows that the intensity of Peak T alone cannot be used to determine bacterial enumeration, especially in systems with complex microbial communities, but supports its use as a measure of microbial activity. The highest intensity for Peak T fluorescence is seen within the resuspended and lysed cells, suggesting that the majority of this material is intracellular, either as structural or functional biological molecules. This explains the presence of Peak T upon inoculation and the increase in intensity with cell multiplication (section 3.1). However, the presence of Peak T in the supernatant also indicates that some of this fluorescence signal is derived from extracellular FDOM, although the amount is species specific, varying from 5 to 25%. This material is possibly associated with metabolic by-products or extracellular proteins (many of which may be functional) that have been exported from the cells.
Peak C fluorescence was observed in both the supernatant and cell lysis fractions for E. coli and B. subtilis (shown in Table 2). Within the supernatant fraction, this can be attributed to either (1) material exported out of the cell (either functional proteins or metabolic by-products) or (2) cellular debris as a result of cell lysis during growth (prior to sampling). However, Peak C fluorescence may also be derived from compounds that fluoresce when not bound within a cell where the fluorescence signal is quenched or inhibited. Peak C is present in all elements of the P. aeruginosa culture, indicating that for this species this FOM is likely to be a functional protein that can be exported to become extracellular DOM. Collectively, this data indicates that the fluorophores that give rise to Peak C fluorescence may be derived from either cell lysis (Elliott et al., 2006) or attributed to microbial metabolic byproducts or extracellular proteins (Guillemette and del Giorgio,  2012; Shimotori et al., 2009).
Peak A C is also seen in all fractions of the P. aeruginosa culture and in the E. coli supernatant. This suggests that this FOM may be a function of a particular biological molecule(s) common to both P. aeruginosa and E. coli. Peak Cþ was also observed in the E. coli supernatant, but at far lower levels compared to P. aeruginosa. The high fluorescence intensity of Peak Cþ in all elements of the P. aeruginosa culture, and the association of this peak with pyoverdine (section 3.1), demonstrates the possible intracellular production and extracellular output of this FOM. Peak M is also observed within all fractions of the P. aeruginosa culture, but is only present in the B. subtilis supernatant. This suggests it may have a similar species specific function like Peak A C , or be derived via the biodegradation of Peak C (Coble, 1996;Coble et al., 2014). As the fluorophores attributed to these peaks (T, C, A C , Cþ and M) can be exported from cells and are identified in cells, lysed cell material and supernatant, they are unlikely to represent cellular structural material. Whilst Peak M is identified in relation to both B. subtilis and P. aeruginosa, Peak A M is only observed in the supernatant of P. aeruginosa, although these peaks have been seen to occur simultaneously in the environment (Coble et al., 2014). Therefore, Peak A M could be attributed to either species specific proteins or bacterial metabolic by-products. From this, Peaks M and A M must be considered separately as they are likely derived from different fluorophores.
Although noted in previous life science research , Peak X (Table 1) has not yet been reported or characterised in aquatic FDOM. However, it is identified at low fluorescence intensities in the supernatant for all species analysed within this study (Table 2). Based on our current understanding of fluorophore structures (Lakowicz, 2006), it is likely that this peak is derived from high molecular weight compounds (characterised as humic and fulvic acids), that would usually be attributed to terrestrial allochthonous material in the environment.
Nevertheless, as it is only seen in the supernatant it is likely to be secreted from the cells and not related to cellular structure.

Future work
The protein-like fluorescence region has been the focus for research investigating microbially-derived, autochthonous dissolved organic matter. The data from this study furthers our current understanding of bacteria-OM interactions and highlights the importance of metabolic activity and bacterial population growth for driving the dynamics of microbially produced FOM and FDOM, albeit within a model system. Furthermore the bacterially derived FOM, exhibits the same fluorescent features as DOM observed in natural systems which has previously been attributed as being allochthonous in origin. This work raises questions regarding the extent to which bacterially produced FOM occurs in freshwater systems and the role that any production plays in the biogeochemical cycling throughout the hydrological continuum. Finally, further work should also explore the metabolic pathways responsible for the microbial production and transformation of FOM and FDOM, including optical regions that are limited by the instrumentation used in this study (e.g. l ex 200e240 nm).

Conclusions
Peak T fluorescence correlates strongly with an increasing bacterial population, but is dependent on microbial metabolic activity. As such, we suggest Peak T as a proxy for microbial activity rather than enumeration. This work provides direct evidence that Peak T fluorescence is ubiquitous within the bacterial cells analysed within this study. It is mainly identified as intracellular material but also exists as extracellular FDOM. Peak C is produced in situ during the exponential stage of bacterial growth curves, likely to be produced via microbial metabolic pathways during microbial growth, or derived from metabolic by-products. FOM peaks can be partially attributed to microbial metabolic processing, through the production of biological molecules, some of which is exported from the cell. These FOM peaks include regions that are currently associated with allochthonous high molecular weight compounds, categorised as humic and fulvic acids. FOM production varies between bacterial species, with this work providing definitive evidence that freshwater FOM can be produced by microbes in situ. It can therefore be of autochthonous origin, altering and enhancing our understanding regarding the complexity of environmental OM origin. Extracellular organic matter contributes to FDOM and, as such, is available as an organic matter source for microorganisms, playing an essential role in nutrient exchange and global carbon cycling.