Aqueous solutions were prepared with ultra pure water (maximum conductivity of 0.055 μS cm^-1) produced by a Sartorius arium pro water purification system (Goettingen, Germany). McFarland turbidity standards were prepared by mixing aliquots of a 1% aqueous barium chloride (Fluka, Buchs, Switzerland) with 1% sulphuric acid (Fisher Scientific, Waltham, MA, USA). A stock solution of 1000 mg L^-1 of bromothymol blue (BTB) (Merck, Darmstad, Germany) was prepared from the dissolution of the solid dye in the appropriate volume of 0.1 mol L^-1 boric acid (Chem-lab, Zedelgem, Belgium) at pH 9.5. The BTB working standards were prepared by stepwise dilution of the stock solution using the same buffer as solvent.

The customized tube holder (Figure 1) was designed using Solidworks 3D software (Dassault Systèmes Solidworks Corporation, MA, USA). The device was manufactured using a Fortus 250mc (Stratsys, Eden Prairie, MN, USA) 3D printer. ABSplus (Stratasys, Eden Prairie, MN, USA), a modified version acrylonitrile butadiene styrene (ABS) thermopolymer, was the material employed. A pair of 6 mm collimating lens (Sarspec, Vila Nova de Gaia, Portugal) with SMA (SubMiniature version A) connectors was placed 20 mm from the bottom of the holder (other distances in the same axis are also possible), and a solid steel platform was assembled to the holder’s base to improve its stability. Complete details about the tube holder parts and dimensions are available in the electronic Supplementary Material.

The detection system unit used in this work comprised a UV-VIS USB4000 miniaturized spectrophotometer controlled through the software Spectra Suite 2.0.162 (Ocean Optics, Dunedin, FL, USA), and a LS-1 visible light source (Ocean Optics). Light was conducted through a pair of 400 μm SMA terminated optical cables (QP-400-VIS-NIR-BX, Ocean Optics). For comparison purposes, a Jasco V-530 UV-Vis benchtop spectrophotometer (Tokyo, Japan) was also used to perform OD measurements in the selected growth experiments. In this case, 10 mm disposable plastic cuvettes were used. OD measurements were performed at 600 nm, and BTB standards were measured at 625 nm. The liquid broth was used as culture blank throughout the experiments.

The pathogenic, facultative anaerobic bacteria Escherichia coli ATCC 25922 and Staphylococcus aureus ATCC 29213 were obtained from the General Hospital of Santo António – Centro Hospitalar do Porto (Porto, Portugal). Both bacteria were revived from -80°C, cultured once on 5% sheep blood agar (bioMérieux, Marcy l’Etoile, France) at 37°C for 24 h and pre-grown at 37°C overnight in pre-warmed Brain Heart Infusion Broth (BHI; Merck, Darmstadt, Germany) with shaking. These pre-grown cultures were diluted immediately before initiating the growth experiments (Harris et al., 2002; Lin et al., 2010).

Strictly anaerobic ruminal bacteria Butyrivibrio fibrisolvens JW11, Butyrivibrio proteoclasticus P18 and Propionibacterium acnes DSMZ 1897 were from the culture collection held at the Rowett Institute of Nutrition and Health (Aberdeen, UK). The type strain of P. acnes DSMZ 1897 was originally obtained from the Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (Braunschweig, Germany) and B. fibrisolvens JW11 and B. proteoclasticus P18 were isolated from sheep (Wallace and Brammal, 1985; Wallace et al., 2006). All transfers and incubations were carried out under O₂-free CO₂ at 39°C in Hungate-type tubes (Hungate, 1969). Inoculum volumes were 5% (v/v) of a fresh culture into 10 mL of medium [liquid form of M2 medium (Hobson, 1969)].

Both pathogenic bacteria suspensions were diluted in BHI in order to obtain a suitable initial OD for the experiments (0.01 to 0.05 at 600 nm). Triplicates of each bacteria were grown in BHI at 37°C in Erlenmeyer flasks shaken at 200 rpm and OD measurements were performed for 9 h with both spectrophotometers (Ocean Optics USB4000 and JASCO UV-Vis, corresponding to a miniaturized setup and benchtop instrument, respectively) until stationary phase was reached (Harris et al., 2002; Baev et al., 2006; Lin et al., 2010). With the anaerobic ruminal bacteria, inoculated culture tubes were incubated in duplicate in a water bath at 39°C. The OD was measured using the experimental manifold at 0, 1, 2, 4, 6, 8, 10, 12, 22, 24, 26, 28, 30, 32, 34, 36, 46, and 48 h.

Bacterial growth curves were analyzed using the grofit R software package (Kahm et al., 2010). For E. coli, S. aureus, and McFarland standards, the relationship between the growth rate values from the growth curves (slope of calibration curve in the case of McFarland standards) obtained from the measurement using either the experimental device or the benchtop spectrophotometer were evaluated by using the PROC MIXED procedure of SAS (version 9.1, SAS Institute, Inc., Cary, NC, USA). Due to the different range of slope values found in each analytical instrument, the mixed model analysis was run with normalized values. The model included the fixed effects of slope and intercept, the random effect of culture, and the random residual error as described by St-Pierre (2001). An autoregressive covariance structure for the intercept and slopes was chosen according to the finite sample corrected Akaike information criterion and the Schwarz Bayesian information criterion (Wang and Goonewardene, 2004). Adjusted observations (slopes) were calculated by adding the residual from each individual observation to the predicted value of the study regression (St-Pierre, 2001). These adjusted observations were corrected for each culture (for the mixed model, McFarland standards were considered a culture).

Considering the objective of developing a simple and versatile system for the measurement of bacterial growth in liquid cultures based on their OD, we designed the tube holder (Figure 1) suitable for the use of commonly used anaerobic culture tubes (Hungate tubes) to perform the measurements. The holder incorporated a pair of collimating lenses (Figure 1) to focus the light beam toward the spectrophotometer. In order to increase the flexibility of the measurements, the lens pair can be moved among three different positions on the vertical axis. The different parts of the holder were 3D printed using ABSplus (additive manufacturing) and then assembled. The optical cables connected the light source to the holder, and the latter to the spectrophotometer.

This new monitoring instrument was used to establish calibration curves using McFarland and BTB standards, in order to individually evaluate the linearity of the scattering and absorption components of the OD measurement (Myers et al., 2013), respectively (Table 1). Hence, we found significant (P < 0.001) linear correlations (r² > 0.99) between absorbance and McFarland standards up to McFarland 8, and a BTB concentration up to 60 mg L^-1. These values corresponded to maximum absorbance values of ∼2.500 in both standard sets. For comparison purposes, the same standard solutions were measured in a benchtop UV-Vis double beam spectrophotometer. Similar linear ranges were observed for McFarland and BTB standards, but with maximum absorbance values of ∼1.500. Therefore, in our proposed miniaturized setup, the sensitivity of the measurement increased 1.8 times for McFarland standards and 1.4 times for BTB standards when compared with the benchtop spectrophotometer. The primary explanation for these differences is related to the optical path used associated with each instrument, which was ∼16 mm for the tubes (proposed setup) and 10 mm for the benchtop spectrophotometer, respectively. This enhanced sensitivity for either absorption or scattering components can impact on the separation of these two absorption components when necessary (Myers et al., 2013), since the miniaturized spectrophotometer is able to perform multi-wavelength measurements.

Bacterial cells may have different light scattering angles according to their size that will directly impact on the OD values recorded (Koch and Ehrenfeld, 1968). Furthermore, the optical geometry of the spectrophotometer, especially optics collimation, may also impact on the measurements, since poor collimation leads to a higher percentage of scattered light that can reach the detector. This corresponds to lower absorbance values that result in linearity deviations of the measurements. In this context it was necessary to perform parallel measurements of OD in our miniaturized optical setup and also in a benchtop spectrophotometer in order to assess the potential use of the proposed methodology for routine measurement of bacterial growth. To this end, each E. coli and S. aureus cultures grown in Erlenmeyer flasks were then transferred to tubes and cuvettes, where OD was measured for nine consecutive hours. As shown in Figure 2, the use of these two optical configurations led to substantially different OD values for the same Erlenmeyer flask cultures of E. coli. Similar differences were observed for the growth curve of S. aureus (data not shown). These differences in the recorded OD values were in agreement with the results observed for the preliminary linearity studies with McFarland and BTB standards (Table 1), and can be primarily justified by the differences on the optical path of each instrument. Therefore, we hypothesized that the differences between the absolute OD values generated by the two optical setups were originated by the differences of the two instruments used, which could affect the measurements due to their different optical paths and/or optical geometry. In order to test our hypothesis, we modeled bacterial growth curves resorting to gcbootspline function (part of grofit R software package), and estimated the three characteristic parameters: specific growth rate (μ), length of lag phase (λ), and the maximum cell growth (A) for both bacterial suspensions measured using both spectrophotometric configurations (Table 2).

Although OD values differed when the two different spectrophotometers were used, the period of time corresponding to the different phases of the growth curves were identical (Figure 2). Thus, we focused on the differences between the μ values obtained using the different optical instrumentation because it reflects the balanced growth of a bacterial culture (Campbell, 1957; Schaechter, 2015). Moreover, it can be used for the comparison of different definitions of bacterial growth measurements (Koch, 2007). Therefore, we compared the normalized μ values of the growth curves of E. coli and S. aureus obtained in both measurements (miniaturized setup and benchtop), as well the slopes of the calibration curves obtained for the McFarland standards under the same measurement conditions. The mixed model analysis showed that the μ values were unaffected by the culture evaluated (P = 0.256). Furthermore, a significant relationship (r² = 0.99, P = 0.004) was also observed between the μ values obtained using miniaturized and the benchtop spectrophotometer [miniaturized spectrophotometer μ = -0.034 (±0.0775) + 1.003 (±0.1183) × benchtop spectrophotometer μ]. These observations are in agreement with the comparative study performed by Matlock et al. (2011), where different UV-Vis spectrophotometers showed different OD values for bacterial cultures and McFarland standards due to the different optical geometries that are present in each instrument. The same authors also described that is possible to calculate a conversion factor between instruments that normalizes data obtained among different spectrophotometers. This was confirmed by the correlation found between the normalized μ values in both instruments used throughout this work. Nevertheless, this conversion factor is a property of each individual culture due to the different absorption and scattering characteristics of the individual cells (Lin et al., 2010; Myers et al., 2013). In the present case, we concluded that similar profiles could be found in both measurement conditions, showing that the miniaturized setup can be adopted as routine analytical tool for bacterial growth profiling.

Another potential application of the measurement strategy proposed in this work concerns the growth of strictly anaerobic bacteria. The growth monitoring of these microorganisms is difficult due to the mandatory manipulation of the culture under anaerobic conditions and the concomitant interference of oxygen. Hence, we evaluated the feasibility of our methodology to monitor the growth of three anaerobic ruminal species - B. fibrisolvens, B. proteoclasticus, and P. acnes – in Hungate culture tubes. After measuring the OD values for 48 h, we modeled the growth curves using the gcFit function (part of the grofit R package; Kahm et al., 2010) that estimated the abovementioned parameters of the growth curves: μ, λ, A (Table 3). In this case, we found a high repeatability between the two different culture tubes prepared from the same inoculum. For all parameters of the growth curves, the relative standard deviation (RSD) values were below 3.5% between the two replicates. This enhanced precision is a key characteristic for the routine profiling of bacterial growth based on the OD of the liquid culture. The setup is also useful in this particular situation, considering that the culture tubes of the anaerobic species studied here must be kept in an O₂-free atmosphere (Stewart et al., 1997).

In UV-Vis spectrophotometers, OD is a light scattering measurement translated by an absorbance value, with the consequence that the use of different instruments commonly leads to different results due to the different optical configurations present (Matlock et al., 2011; Myers et al., 2013). Typical UV-Vis benchtop spectrophotometers for OD measurement comprise microplate and dual-beam formats that usually ensure a suitable sensitivity and linear range. However, they also require the transfer of the liquid culture to a microplate or a cuvette or a dilution protocol (Matlock et al., 2011; Myers et al., 2013), which presents several constraints when pathogenic or anaerobic bacteria have to be measured. In contrast, the setup proposed here was able to bypass sample handling when the bacteria are directly cultured into the tubes (this was the case with the rumen bacteria), and can also be used in appropriate places for sample manipulation (e.g., laminar flow hoods) due to its portability. Additionally, it is also possible to measure OD at different wavelengths and/or the complete absorption spectrum of the sample. This also impacts in the growth profiling of species or media that have a strong absorption component (Matlock et al., 2011; Myers et al., 2013) or when other light scattering sources may be present (Hernández and Marín, 2002; Bernardez and De Andrade Lima, 2015).

Following this validation of the novel instrumentation, future development could take advantage of further developments in 3D printing solutions (Gross et al., 2014), LED light sources (Macka et al., 2014), and flow cell designs (Tymecki et al., 2008; Myers et al., 2013), which would lead to new online monitoring tools for the accurate and precise measurement of bacterial growth.