The signal intensity obtained from the stained cells can be improved significantly by performing the detection of the fluorescent signal at certain excitation and emission spectra. The optimal excitation and emission spectra are dye dependent and can furthermore be influenced by environmental conditions such as salinity [11] [13] [16] .

A detailed analysis of the excitation (EX) and emission (EM) spectra for PicoGreen and SYBR Green dyes was performed to determine which spectra should be implemented in the sensor prototype to obtain the lowest possible detection limit. A high-sensitivity 920 FLS spectrofluorometer was applied for the experiments. The tests were performed in seawater from Aarhus bay spiked with NaCl to match the salinity of Saudi Aramco injection seawater. An example of the test results is presented in Figure 2, which shows the variation in the signal intensity for the PicoGreen and SYBR Green

stain, when different EX and EM spectra were applied. The optimal EX and EM spectra for PicoGreen and SYBR Green dyes were determined to be EX490 - 505 nm/EM 518 - 523 nm and EX490 - 498 nm/EM 518 - 522 nm, respectively. The optimal spectra were used for the optimization of DNA staining and for the selection of sensitive and highly specific detector implemented in the sensor prototype.

SYBR Green is a cell permeable DNA stain, while PicoGreen cannot penetrate the cell membrane and needs cell permeation treatment prior to staining to increase DNA binding efficiency and thereby improve the bacterial detection limit [11] . Several permeation treatment methods [11] [13] - [15] [23] , including ionic and nonionic detergents (sodium dodecyl sulfate or SDS, and Triton X-100), formaldehyde fixation, and cell destruction by liquid nitrogen, have been tested with enrichment culture grown in artificial seawater media with 5.5% salinity and pH 6.8 to mimic the conditions in Saudi Aramco injection seawater system.

The signal intensity for PicoGreen could be significantly increased by either destroying cells with liquid nitrogen or treating cells with Triton X-100 (Figure 3(a)). SYBR Green staining showed equally high signals with destroyed cells and non-treated cells (Figure 3(b)). This illustrates that SYBR Green is able to penetrate efficiently into the living cells without any pre-treatment. Triton X-100 method was excluded from further testing as this viscous material is likely to cause tubing clogging in the sensor and erroneous measurement of the cell density caused by air bubbles formed during mixing [24] . SDS pre-treatment and formaldehyde fixation were also excluded, as they did not improve the fluorescent signal. The treatment marked with green in Figure 3 was selected for further experiments to determine the detection limit of DNA staining technology.

The detection limit of the DNA staining technology was tested for PicoGreen and SYBR Green stains, applying the optimum excitation and emission spectra and the optimum pre-treatment of the cell sample. The detection limits were determined by applying the staining technologies on a serial dilution of bacteria in seawater with 5.5% salinity.

Both SYBR Green and PicoGreen stains had detection limits down to 2 × 10³/ml without any preceding treatments of the cells (Figure 4 and Figure 5). This confirms that the optimized method for these stains will be sufficient to detect the low cell numbers in the Saudi Aramco injection seawater system. The signal intensity for SYBR Green (Figure 4(b)) was higher at low cell concentration than PicoGreen (Figure 5(b)), and therefore, SYBR Green was evaluated to be more reliable for detection of the low cell number found in Saudi Aramco injection seawater.

When applied on destroyed enrichment culture grown in artificial seawater media, the detection limit of SYBR Green and PicoGreen was 1 × 10⁵ cells/ml and 1 × 10⁴ cells/ ml, respectively (Figure 6 and Figure 7). The higher detection limit was very likely caused by high background fluorescence from residual DNA of yeast extract [25] ,

which had been added to the artificial growth medium to stimulate bacterial growth. Therefore, the detection limit obtained in seawater will presumably best represent the detection limit in Saudi Aramco injection seawater.

The detection limit of 2 × 10³ cells/ml for SYBR Green and PicoGreen without prior treatment will be sufficiently low to detect the increases in biomass in Saudi Aramco injection seawater.

Pre-treatment of sample may increase signal intensity and sensitivity of the sensor technology, but will also complicate sensor construction. Several chemical and me- chanical pre-treatment methods were evaluated for SYBR Green and PicoGreen dyes (Figure 3). The results indicated that SYBR Green can penetrate into the living cells efficiently without any pre-treatment with a low detection limit of 2 × 10³ cells/ml (Figure 4(b)). PicoGreen also achieved the same detection limit without any preceding treatments of the cells; however, signal intensity was lower at low cell concentration than SYBR Green (Figure 5(b)).

The detection limit of destroyed cells with PicoGreen and SYBR Green was 1 × 10⁵ cells/ml and 1 × 10⁴ cells/ml, respectively, higher than the detection of non-treated cells with corresponding stains (Figure 6 and Figure 7). In addition, cell destruction procedure requires energy and will significantly complicate the construction of an online sensor.

SYBR Green staining of non-treated cells is therefore the recommended technology for application in the automated sensor prototype.

Larger particles can interfere with fluorescence emission from the cells [26] . The particles interfere by causing scattering of the fluorescence signal, resulting in decreased detection limit. Pre-filtration at the sampling inlet of the sensor removes larger particles from the sample, but complicates automation of the technology, as filters may need frequent replacement due to clogging and/or bacterial growth on the filter surface. The need for incorporating a filtration step in the DNA staining technology was evaluated.

The Saudi Aramco injection seawater contains a low number of particles larger than 2 µm (approximately 600 - 1000 particles/ml). The majority of particles observed in the 2010 survey were very small, with only 10% of the particles being larger than 2.89 µm. Raw Danish seawater contains markedly more particles than Saudi Aramco injection seawater, and can be reckoned as a worst-case scenario. Filtration of raw Danish seawater samples (10-µm filter) did not influence the standard deviation of fluorescence readings (16 samples) compared to non-filtered seawater (Figure 8). The fluorescence reading before and after filtration was approximately the same. It is therefore concluded that pre-filtration of Saudi Aramco injection seawater was not necessary before bacterial quantification with DNA staining technology.

A simple automated microbe sensor can be designed and constructed based on the SYBR Green staining technology, where a low detection limit is achieved without

chemical or mechanical pre-treatment of the injection seawater sample. Furthermore, automation of the staining procedure can be further simplified by omitting sample filtration prior to analysis.

The fluorescent dyes may lose stability during prolonged storage at elevated temperatures [13] [14] . The demand for temperature regulation in the dye storage compartment of the sensor was tested by incubating the stain at different temperatures for prolonged time (26 days for SYBR Green and 30 days for PicoGreen).

SYBR Green storage stability was tested with concentrated and pre-diluted (1:10) dye at −20˚C, 5˚C, 20˚C and 30˚C. PicoGreen was incubated at 5˚C (recommended storage temperature), 20˚C and 40˚C. The fluorescent signal intensity of the concentrated SYBR Green was very stable and independent of storage temperature (Figure 9(a)). The signal intensity after staining with pre-diluted SYBR Green dye was more fluctuating (data not shown). We therefore recommend that SYBR Green is applied in the sensor prototype in its concentrated form, and that the solution is stored below 30˚C.

The signal intensity of the PicoGreen stain decreased gradually during 30 days’ storage at elevated temperature. The results indicated that storage at 5˚C is required for PicoGreen, to avoid a false decrease in detected cell densities, due to dye degradation (Figure 9(b)).

The temperature can affect the fluorescence signal intensity either by influencing dye stability or by causing dissociation of DNA strings and thereby hindering dye binding to cellular DNA [9] [10] [13] [14] . The need for temperature regulation during detection of the fluorescence signal was investigated by measuring the fluorescence intensity at different temperatures with 5˚C intervals from 25˚C to 65˚C on seawater and sterile seawater.

The fluorescence intensity of both PicoGreen and SYBR Green dyes was highly influenced by temperature during signal detection (Figure 10). Fluorescence signal from SYBR Green remained stable at temperatures between 25˚C and 35˚C, while PicoGreen fluorescence decreased as soon as the detection temperature increased from 25˚C. Due to the high temperature fluctuations expected in the Saudi Arabian dessert, it is re- commended to incorporate temperature regulation in the chamber of the sensor prototype where the fluorescence signal is detected.

Temperature variations may influence binding of the dye to the cellular DNA and thereby influence signal intensity obtained [9] [10] . The need for temperature control during sample staining was tested by comparing the fluorescence signal obtained from cells, which had been incubated with stain at different temperatures.

PicoGreen fluorescence intensity was independent of whether the sample was incubated at 22˚C or 30˚C, while the fluorescence intensity for samples incubated at 37˚C decreased slightly (Figure 11(a)). The signal intensity from SYBR Green stained samples incubated at different temperatures showed no clear-cut pattern (Figure 11(b)). Overall, the fluorescence intensity was only slightly influenced by the incubation temperatures during staining.

Temperature control is needed for obtaining stable fluorescence signal intensity with DNA staining technology, if the sensor is required to take fluorescence readings at any given time. SYBR Green signal intensity was not influenced by elevated temperatures up to 30˚C, and SYBR Green is therefore again the better suited dye for the sensor prototype.